This PDF file contains the front matter associated with SPIE Proceedings Volume 12330, including the Title Page, Copyright information, Table of Contents, and Conference Committee listings.

With the gradual improvement of modern enterprise internal control, the comprehensive risk management system plays a more and more important role in ensuring the sustainable development of enterprises, promoting the construction of comprehensive legal system, strengthening enterprise risk control and management, and improving the level of enterprise risk prevention. This paper summarizes the fuzzy clustering theory, risk management and AFDD algorithm, puts forward the risk identification management system, and analyzes the risk evaluation results. The results show that H company analyzes database permission risk (0.357), platform and system permission risk (0.637) and database account risk (0.725). These three risks exceed 0.1, which belong to the category of high risk and need to be paid attention to and solved first.

People will analyze the faithfulness of others through facial features and expressions when they come into contact with strangers. In mating choice, the accurate estimation of sexual unfaithfulness will reduce the reproductive cost of an unfaithful mate. Therefore, this paper uses machine learning to judge sexual faithfulness based on human perception of the face in four dimensions: attractiveness, trustworthiness, faithfulness, and sexual dimorphism. First of all, this study uses three conventional machine learning algorithms to classify the data. Then we use soft voting to build an ensemble model. The study reveals that women have better performance than men at judging sexual faithfulness. Moreover, the AUC of our ensemble model is 0.861 based on female and 0.789 based on male.

Reducing the unemployment rate has become a serious social problem facing the world. In this article, we used the average wages of private sector employees in cities by industry and the entropy of these wages as characteristic variables to analyze the unemployment rate of 30 major provinces and cities in China from 2011 to 2019. We use K-Nearest Neighbors (KNN), Support Vector Machines (SVM), and Adaptive Boosting Algorithms (Adaboost) to classify areas with high and low unemployment rates. Then we perform linear regression analysis based on the results, analyze the correlation between average wages and income inequality, and interpret the classification results according to the decision boundary. In conclusion, we find that in regions with low unemployment rates in China, higher average wages are often accompanied by greater income inequality, while in regions with higher unemployment rates, the situation is more moderate. Compared with areas with low unemployment rates, the increase in average wages in areas with the same high unemployment rate has brought about a smaller increase in income inequality.

Traditional single-model text feature representation methods lack multi-level representation of features and cannot extract global and local information well. To this end, the text uses a two-layer ReGU network (BiReGU) and a capsule network to extract the global semantic and local semantic features of the text respectively, combining attention mechanisms and interaction forms to obtain a more comprehensive multi-level feature representation of the text. In addition, the traditional capsule network dynamic routing algorithm suffers from the problem of noisy capsule interference. This paper proposes to reduce the weight of noisy capsules based on the attention mechanism and Capsule's feature extraction method, thus reducing the interference to subsequent capsules. Comparative experiments are conducted on a generic dataset and the experimental results demonstrate the effectiveness of the proposed method.

In recent years, the incidence of cardiovascular disease has been on the rise, and the causes have gradually become more complex, so it is urgent to intervene in the diagnosis and treatment of heart disease. In this paper, 251,505 survey responses from BRFSS were selected to analyze the cardiology dataset using the Logistic model and the C5.0 decision tree model. The results show that the occurrence of heart disease has a strong relationship with hypertension, smoking, stroke history, gender, and age, which has certain practical significance for the prediction and diagnosis of heart disease.

With the development of speech synthesis technology, the simulation of specific individual’s speech has gradually matured， synthetic speech is easily and perceptually recognized as real speech, which may occur frequently in illegal activities. To identify crimes, forensic technology is widely used such as comparing the formants, pitches, and rhythm. The present study aims to investigate whether the method of comparison of formants can recognize the differences between the perceptually similar synthetic speech (hereinafter “personal anchor” speech) and real speech. To this end, two young males and two young females from different dialect regions were recruited to read the same text. Their voices were recorded and used to generate four “personal anchors” by the software of sound spectrum and statistics analysis. The method of comparison of various parameters of the formant, including numerical statistical, stability analysis, and transitional segments feature were applied to analyze the differences between the real speech and the corresponding “personal anchors”. It was found that the numerical or stability analysis of formants was not sufficient to fully determine whether the speech was synthesized, while comparing the transitional segments of some specific syllables could efficiently detect the synthetic speech from the real speech.

Random numbers are the security foundation in the field of cryptography and information security, providing the most basic security guarantee for cryptography and security protocols. The quality of random numbers depends to some extent on the randomness of the data generated by the random number generator. For devices such as mobile smart terminals, high-quality true random numbers cannot be generated because a true random number generators cannot be embedded. However, these usage scenarios have certain requirements on the quality of random numbers. A random number scheme, that is, generating higher-quality random numbers without increasing hardware investment. In this paper, a software random number generator based on the Shangmi algorithm is implemented for the situation where there is no true random number generator. The existing hardware or components are used as the entropy source, and the random information generated by the entropy source is collected as the initial data, and processed through the Shangmi algorithm. Generate a random sequence, and the final generated random number meets the Shangmi detection standard.

The use of malicious loading like ROP to carry out remote vulnerability attacks has become a severe concern to computer systems. It is critical to reliably detect ROP assaults in traffic and to successfully detect and block remote attacks before they have an impact on the attack target. Existing traffic-based detection approaches, on the other hand, rely on searching for particular bytes in the traffic, which has a low detection efficiency and a significant cost. This study proposes a lightweight ROP attacktraffic detection approach based on image representation, which can identify the ROP chain's image representation from network traffic and therefore complete the detection. We extract gadget fragments from network traffic using ROP gadgets of genuine attacks and randomly mix them with regular traffic to create a training dataset. The inputs are then fed into a Convolutional Neural Network (CNN). Our tests demonstrate that our accuracy reaches 99.7%.

With the popularity of cloud storage technology, mobile devices can store or retrieve personal information more and more conveniently. However, the data in cloud storage faces more and more security risks, which limits the development speed of cloud storage technology. In order to promote the development of cloud storage technology, it is very important to solve the security problem. At present, there are many solutions for cloud storage security, but most of them do not work well in the mobile cloud storage environment. Therefore, this paper studies a secure sharing scheme of mobile cloud storage data based on attribute encryption. This scheme uses CP-ABE as the access control technology, which has been proved by experiments to ensure the security of user data in mobile cloud storage.

As a new board independent from the main board market of the stock exchange, the Science and Technology Innovation Board with registration system is of great significance in improving the basic system of the capital market in China. China's IPO underpricing phenomenon in the stock market plate is very prominent, but rare research has been done in studying the phenomenon of IPO underpricing in science and technology innovation board.

Massive collection and utilization of data by artificial intelligence systems makes data analysis automated and intelligent, and the use of intelligent algorithms poses challenges to privacy and data protection. Disputes over data ownership have brought difficulties to the legal and safe use of data. By analyzing the existing data ownership views and laws in China, the European Union and the United States, it is believed that the existing legal system cannot bring effective supervision and management to the use of data. The way to create a data trading platform to improve the existing management and supervision mechanism is to feasible mode.

At present, China's auto insurance pricing is relatively rough, without considering the vehicle type factor, and the use of a generalized linear model has many limitations, resulting in the low accuracy of auto insurance rating. In order to solve the above problems, this paper integrates the vehicle physical collision test results and other vehicle factors into the automobile insurance pricing model. Based on the vehicle physical parameters and vehicle physical collision test data, claims data, and underwriting data, this paper establishes the GLM and the cutting-edge machine learning model LightGBM respectively and compares the automobile insurance pricing methods. The findings reveal that the new model component incorporated into the pricing model has a significant impact on the final vehicle damage insurance total claim. In addition, the prediction accuracy and model robustness of the machine learning model LightGBM in dealing with the entire amount of car damage insurance claims with numerous features and huge data are higher than those of the classic vehicle insurance pricing model GLM. This paper lays out a workable plan for achieving a fine vehicle insurance pricing that is both fair and unique.

In terms of network security visualization, due to the diversity and complexity of security data, visualization technology can be used to analyze the security of the entire information system, detect intrusions, investigate network traffic, and verify the rules and policies of security sensors. The traditional prediction method is difficult to accurately model the relationship between the situation value and the time change. The prediction error is large. This paper proposes a cycle gating unit coding prediction method based on the attention mechanism. The GRU network is used as the basic neural network processing unit. The attention mechanism is introduced into the traditional GRU network layer to calculate the attention weight of different safety data. After data fusion, the GRU network is input to predict the situation value. Simulation results show that the proposed method has smaller prediction error, faster convergence speed, and lower complexity. The architecture proposed in this paper has a good dimensionality reduction effect and classification accuracy of situation elements, achieves excellent cyberspace situational awareness, and improves network security through the visual operation of cyberspace.

At present, privacy issues have become one of the great obstacles to the further development of web technology. If it cannot ensure that privacy is protected, then all kinds of user data and related privacy can be stolen, which will seriously hinder the trust of individuals, enterprises and others in technology, and thus the progress of technology. Therefore, privacy protection technologies are of particular importance. Through the analysis of Internet security incidents, this paper explains the research background and significance of privacy protection, and analyses the development status of privacy protection and summarises the challenges faced by privacy protection; then by analysing blockchain technology, it makes a research analysis on the feasibility of its application to privacy protection, gives the research ideas and specific research contents of this thesis, and finally describes the organisational structure of this thesis. The study provides new ideas for subsequent research in this field.

Anti-counterfeiting is very important in commodity circulation area because it protects the interests of producers and consumers. Product identification is a kind of common anti-counterfeiting means. An innovative method was developed in this paper with aim to identify packaged products by means of printed image. Target information was hidden in digital gray patch by use of this method and then the digital gray patch was duplicated through digital printer to achieve a hard copy. The information hidden was invisible on the hard copy and could be extracted via a decoding software developed in this paper. The method created here is radically different from the digital watermark technology, since it intends to hide information in hard copies that usually adopted by commodity packages. Experiment results showed that various information could be hidden well in printed duplicates and identified clearly through the decoding software. The method constructed by this paper is low-cost and effective and therefore provided a new candidate for the field of commodity security and identification.

Fog computing is an evolving paradigm which extends cloud computing to the edge of the network. Recently, how to combine the fog computing with cloud computing will play a significant role in Internet of Things (IoT), smart grid, health care, etc. However, fine-grained data access control and quantum computing attack resistance are two main secure problems in fog and cloud computing environments. To handle these problems, in this paper, we propose a multi-authority threshold attribute-based encryption scheme from R-LWE in fog and cloud computing environments (MAT-ABE-FC). Our MAT-ABE-FC scheme innovatively integrates the traditional cloud computing with the promising fog computing environment, and can not only provide fine-grained data access control, but also withstand quantum computing attacks. Moreover, our MAT-ABE-FC scheme can support multi-authority, and realize secure attribute revocation. Both security and performance analysis indicate that our scheme is secure and efficient.

Recently, Blockchain technology has been widely used in food traceability service. However, several problems are still not properly solved in the current schemes, such as the inconvenience of management, the lacking private information protection of participants and the poor performance in external traceability service. In this paper, we propose a novel food traceability Scheme based on multi-tier cooperation consortium blockchain. In our approach, business and data are managed by hierarchical architecture, private information of traceability participants is effectively protected and performance of external service is improved at the same time. Experimental results based on Hyperledger Fabric framework validate the proposed solution.

The Correct identification of entities in military requirements documents is an important basis for constructing military requirements graph and realizing information-based operations. Aiming at the problem that the existing named entity recognition methods cannot effectively solve the problem of inaccurate identification of complex named entity boundaries in requirements documents, a small data sample of military requirements documents is studied, and a named entity recognition model based on BERT and global pointers is proposed (BERT-GloP-Rule). The model first uses BERT to encode to obtain deep text semantic information at the word segmentation level and sentence level, and then uses the global pointer to decode, treats the beginning and end of the entity to be identified as a unified whole to discriminate, and finally combines rule matching to complete entity-level semantic information. Requirements document named entity recognition. The experimental results show that the precision of the model on the test set reaches 90%, the recall rate reaches 89%, and the F1 value reaches 91%, which verifies the recognition effect of the BERT-GloP-Rule model on complex entities in the field of military requirements. Significantly better than traditional machine learning models.

Accurate power grid investment can not only make the control of power grid construction more precise, but also improve the level of investment decision-making. Firstly, by constructing the investment efficiency model of power grid business, this paper solves the qualitative limitations and economic one sidedness of traditional power grid investment evaluation. Secondly, based on the research on the theory of power grid investment efficiency, the calculation model of power grid enterprise investment efficiency is proposed. Thirdly, through example analysis, the power grid investment efficiency of cities under different power grid services is calculated. It also analyzes the horizontal and vertical reasons behind the investment efficiency, and verifies the rationality and effectiveness of the investment efficiency model of power grid enterprises. Finally, its application scenario is proposed, which lays a certain foundation for continuously promoting the construction of high-level power grid.

Monocular vehicle perception has been a vital problem in autonomous driving. Anchor-based detectors are always used to solve the problem but the used bounding box’s center to calibrate is actually not corresponded with vehicle’s 3D center. Therefore, we designed a perception network by point detector U-Net and propose a calibration method by using the 3D2D reprojection constraints and a prior plane in the imaging system. We evaluate our method on the ApolloCar3D dataset, and the experimental results prove that the proposed network is feasible and the calibration is effective. On the ApolloCar3D testing dataset, the mean average precision of the proposed network is 0.022, and can be well improved to 0.040 with our calibration method, which is competitive compared to other algorithms.

With the improvement of people's quality of life, the number of cars in China is gradually increasing, and the auto insurance industry is developing rapidly and is an important part of property insurance. However, China's auto insurance market is still far from the international level in terms of accuracy rating determination: the international mature insurance market generally adopts the car-based pricing model, and the development of this model in China is relatively backward. Generalized Linear Models (GLMs) are commonly used to predict the intensity of auto insurance, but they require high data requirements and insufficient samples, which may lead to unreliable parameter estimates. In this paper, the Generalized Additive Models for Location, Scale and Shape (GAMLSS) can improve the limitations of GLM and perform better on various distribution parameters. In addition, China currently focuses on the purchase price of new cars as the main rating determination factor, which fails to match the differences of different claims caused by the crashworthiness and maintenance economy of the car itself with the motor vehicle premiums. In order to fill the gap in this area in China, this paper will introduce the results of crashworthiness and economic maintenance of China Insurance Automotive Safety Index (C-IASI). And machine learning-random forest and PAM are used to reduce the dimensionality of the independent variables to obtain the composite variables, after which the fitting effects of GLM and GAMLSS with the addition of machine learning factors are compared. In this paper, we use measurement results of C-IASI, multi-parameter GAMLSS, random forest and PAM to increase the accuracy of auto insurance pricing.

The paper is going to price snowball autocallable by Binomial Model, which is unconventional as the product is usually priced by Monte Carlo Simulation and Partial Differential Equations (PDE) Method. Using Snowball VWO issued by Barclays Bank as an example, we set up three datasets of the underlying ETF prices and obtain three volatilities. For each dataset, we replicated three portfolios based on the payoffs on three call valuation days and computed the sums of these three portfolios to be the predicted prices. The pricing results shows a strong correlation between volatility and the price of snowball autocallable. However, by comparing these theoretical prices with the price stated in the pricing supplement, we find that the snowball autocallable is overvalued in reality. The results in this paper benefit pricing researchers who attempt to use Binomial Model for pricing snowball derivatives and investors who are interested in snowball autocallable.

As a product of the market economy, the stock market plays a crucial role in the financial field of a country. And gold as an important choice for hedging and maintaining value, it has also attracted wide attention. To help traders make the best decisions based on historical data, this paper establishes prediction and decision models for gold and equity data. Firstly, the time series prediction model Autoregressive Comprehensive Moving Average model (ARIMA) is used to analyze the historical data and predict the results of the next day. After that, based on the principle of maximizing returns, a decision model is established based on dynamic programming (DP) to develop the best portfolio for investors. The sensitivity of the transaction cost of the decision model is analyzed according to real transaction rules. It provides a more accurate model for the relationship between variables in the financial market. The quantitative investment decision model established in this paper has certain practical value and can provide a certain reference for investors in the financial market.

Disinformation has brought serious harm to the information ecology. The key point is how to mitigate it using disinformation verification. In this regard, this paper systematically analyze the importance and structural relationship among the influence factors of the online disinformation verification environment. It helps to deepen the understanding of online disinformation verification environment and better promote the realization of online disinformation verification. Based on the literature and interviews, 13 influence factors of online disinformation verification environment are identified by Delphi method. Using the matrix scoring questionnaires of experts and related personnel as samples, the Decisionmaking trial and evaluation laboratory method was used to analyze the relevant associations between the attributes. The results show that the level of network economic development, information literacy level, institutional implementation environment, education level, information analysis technology, information dissemination technology, and information collection technology are the key factors influencing the online disinformation verification environment.

With the development and progress of electric power communication technology in China, it gradually covers and improves the diversified and personalized needs of users. Different from the traditional network construction, the construction of a soft switch network is more systematic and advanced to better serve the development of network telecommunications and communication systems. This paper is based on the SIP protocol control technology as the core of the soft exchange. Let the network business media flow direct bear on the IP network through the soft exchange system and control and transfer in a business organization structure, which is to make the application layer control mainly in the business running on the server, the data, voice, video, news and other business processes. Through the experiment, the maximum number of registered users and the call processing ability of users in the domain of the design scheme are tested and verified, which proves that the scheme in this paper is superior to the traditional voice bearer method of the power grid to meet business needs. It can provide technical support for power grid production and lay a solid foundation for the sustainable and healthy development of the power grid in the future.

In the fields of information security, the application of random numbers is becoming more and more widespread. True random numbers (TRNs) tend to have better randomness and unpredictability than pseudo random numbers (PRNs). Chaotic systems are very suitable for constructing true random number generators (TRNG) due to their periodic infinity, randomicity, long-term unpredictability, high sensitivity to initial values and parameters, etc. However, when chaotic systems are implemented by finite state machines, their dynamic characteristics will disappear, such as periodic infinity and initial value sensitivity, and there are serious safety concerns about the random number generator designed on this basis. In order to address the degradation problem of the chaotic systems, this paper proposes a true random number generator based on the chaotic circuit, implements a complete hardware system, and tests its randomness.

The insulation structure research and design optimization of ±1100kV DC SF6 gas insulated through wall bushing is an important basis and technical key for the core technology research and device development of ± 1000kV DC SF6 gas insulated through wall bushing. It is also an important basis for the formation of bushing structure design and process technology, which is of great significance to improve the electrical performance of bushing. The ±1100kV DC SF6 gas insulated through wall bushing monomer bears the full voltage and current of the system, and the distribution of internal insulation electric field is complex. At the same time, it is affected by many factors such as electric field shielding, insulator surface charge accumulation, material non-linearity and so on. The improvement of voltage level puts forward higher requirements for bushing structure type, material performance and electric field uniformity regulation. Based on the actual engineering and scientific needs, this topic carries out the insulation structure research and design optimization of ±1100kV DC SF6 gas insulated through wall bushing. The design and selection of the bushing insulation structure, high current carrying structure and mechanical support structure based on multi physical field analysis. Through electrical, thermal and mechanical analysis, design and select reasonable structural type for casing under the action of multiple factors. Design optimization of typical insulation structure of bushing. The electric field simulation and intelligent algorithm are used to design the configuration parameters of shielding electrode and epoxy cast insulator, optimize the electric field intensity on the electrode surface, inhibit the charge accumulation on the insulator surface and improve the flash-over voltage of insulator.

With the development of the Internet, Linux has received support from software enthusiasts, organizations, and companies all over the world. In addition to maintaining a strong development momentum in servers, it has made great progress in personal computers and embedded systems. But the popularity of Linux has also made it the target of numerous hackers. Linux backdoor not only seriously affects people's daily experience, but also threatens users' property safety, and even directly threatens the normal operation of society. In this paper, the N-Gram model is used for processing, and the TF-IDF algorithm is used to further improve the classification performance, and the ensemble learning algorithm is used to train a high-performance Linux backdoor classification and recognition model. Through experimental analysis, the accuracy of the model is over 97%. Compared with conventional classification models, this model has superior performance.

In order to improve the accuracy of fine-grained image recognition, a fine-grained image recognition method based on integrated transfer learning is proposed. This method uses two models of VGG-16 and inception-v3, combined with transfer learning to construct a K-selection transfer learning algorithm. The network model parameters pre-trained on the large-scale natural image dataset ImageNet are transferred to the transfer framework model of the fine-grained image dataset, and the network model is fine-tuned. The algorithm proposed in this paper is tested on CUB-200-2011 dataset, Oxford 102 Flower dataset, Stanford Cars dataset, FGVC-Aircraft dataset and Stanford Dogs dataset. The experimental results show that, compared with existing algorithms, the proposed algorithm has improved fine-grained image recognition accuracy under different models, and has optimized the adaptive effect of migration.

Wireless Body Area Network (WBAN) can be mainly applied in medical field for monitoring patient physiological information. However, information leakage is hard to completely avoid during data transmission via wireless communication links. Therefore, it is of great practical significance for the research on the authentication protocol in WBAN. In this paper, we propose a WBAN authentication protocol in a multi-gateway environment. The proposed scheme allows mutual authentication between data user, gateway and patient device which gathers data from sensors. Once the mutual authentication is successful, data user and patient device can agree on a secret session key for subsequent communication. In particular, we introduce an asynchronous method to achieve forward security. We prove the security of the scheme based on real-and-random model and via informal security analysis. Finally, the performance comparison with similar schemes shows that the proposed scheme is efficient in terms of computation and communication costs.

This article in view of the user's location information security sharing scheme analyzes the core technology and common sharing model, combined with the user's location information security core sharing scheme, including initialization processing, recording, location sharing, system security, etc., and its aim is to increase the use value of the sharing scheme, fully protect the user's information security.

In order to strengthen the power network structure of our country, reduce the frequency of electrical equipment and lines, so that the power network can operate normally and smoothly. Based on the above reasons, the flow feature association mining algorithm is introduced into the research of power network security situation assessment to evaluate and predict the safe running state of power network. Firstly, this paper explains the basic algorithm of stream feature mining, including time correlation, IP (Ingress Protection) correlation, type correlation and super warning correlation. Secondly, it is necessary to determine the weight of network security evaluation index and the calculation method of evaluation, and carry out the relevant case analysis. Finally, the network security situation assessment model based on the flow feature association mining algorithm was compared with the BP(Back Propagation) neural network and RBF(Radial Basis Function) neural network evaluation model, and the comparison results showed that Flow feature association mining algorithm can improve the feasibility and accuracy of power network situation assessment. In conclusion, the study of this paper can provide a new research approach for the construction of network security situation assessment.

In recent years, the innovation capability of the Pearl River Delta region has undergone significant changes. The paper, based on cluster analysis and factor analysis, seeks to analyze the development level, differences, and influencing factors of the innovation capability of the central cities in the PRD in 2020. The results indicate that those factors driving innovation include spending on research, science and technology profit, science and technology output, research potential, education, and consumption level. Based on the scores of the three factors, Shenzhen ranks first with an overall score of 1.233, while Zhuhai ranks last with an overall score of -0.468. The development of innovation capability in the PRD city clusters forms an innovative city network with Guangzhou and Shenzhen as the core, Foshan, and Dongguan as the subcenters, and other cities such as Jiangmen, Zhaoqing, Huizhou, Zhuhai, and Zhongshan as the third echelon. To enhance the overall innovation capacity in the PRD, an innovation community must be built based on the integration strategy of the PRD region. Greater efforts must be made to enhance the innovation capacity in three aspects: spending on research as well as science and technology profit, science and technology output together with research potential, and education and consumption level.

In order to solve the problem of high false positive rate of regional economic big data anomaly detection, a regional economic big data anomaly detection method based on time series algorithm is designed. Select regional economicindicators and divide socio-economic statistics and spatial data; Optimize the data space conversion process, select influencing factors and modeling methods; The simultaneous equation model is established by using time seriesalgorithm, and the anomaly detection mode of big data is established to realize the anomaly detection of regional economic big data. Experimental results: the average false positive rates of the regional economic big data anomalydetection method proposed in this paper and the other two regional economic big data anomaly detection methods are2.477%, 4.060% and 3.986% respectively, indicating that the detection method designed in this paper has lowfalsepositive rate and better detection performance.

With the deepening and strengthening of the application of computer tools, wireless signal technology and storage technology in the field of financial accounting, accounting computerization has excellent performance in the high efficiency and accuracy of financial data processing. A lot of excellent algorithm formula also comes from the accounting computerization. This paper analyzes the balance sheet and fixed assets investment table, and reveals the relationship between the internal elements of the statement and the relationship between the statements. This paper describes the relationship tree in an image way, and marks the corresponding calculation algorithm, and completes the flow chart of data adjustment and data balance with general meaning in the data calculation of accounting computerized statements, and verifies its correctness from the scientific point of view. This article studies the accounting computerization data adjustment method based on artificial intelligence algorithm, introduces the accounting computerization data adjustment method related content and knowledge. Through data testing, the results show that the research of accounting computerization data adjustment method based on artificial intelligence algorithm greatly promotes the efficiency of accounting computerization data adjustment.

A fast and accurate fish identification system is urgently needed for the development of fishery industry. In this paper, ten kinds of common fish are taken as the research object, and the existing achievements of deep Convolutional Neural Networks in the field of image classification are used to provide technical support for the establishment of intelligent and efficient fish recognition system in fishery production. The research mainly includes three parts : data collection, model building and optimization comparison. Firstly, fish images are collected through web crawler and field collection, and then the collected image format is normalized to establish a fish data set. data augmentation is used to expand the training dataset and reduce the dependence of the model on the original dataset. Secondly, AlexNet, VGGNet and ResNet convolutional neural networks are used for image classification tools, and the generalization ability of the model is improved by combining the transfer learning method. Finally, through the comparison of the training results of each model, the ResNet-50 network model with better performance based on transfer learning is found as the main model of fish recognition, and its recognition rate is 98.06 %.

In order to solve the problem of slow convergence and local optimum of traditional crisscross algorithm, a hybrid topology crisscross algorithm (HTCSO) is proposed and applied to solve the dynamic economic dispatch problem of power grid. The hybrid topology combining ring regular topology and stochastic topology is adopted to increase the diversity of the population and expand the solution search space, so as to improve the convergence accuracy and convergence rate of CSO algorithm. Experiments on small-scale 10 units system and large-scale 100 units system show that the proposed hybrid topology crisscross algorithm has faster convergence speed and better convergence accuracy compared with other swarm intelligence optimization algorithms, which provides a reference case for real-time accurate dispatching and carbon emission reduction of power grid.

Credit risk management, which is the basis of the credit application, is the most perfect embodiment in the bank credit application and asset supervision. The ultimate purpose of credit risk management is to ensure that credit fund is of safety, profitability and fluidity. At present, it is extremely important for commercial banks to set up an early bank risk warning system. Therefore, this paper considers the use of six kind of machine learning methods to model credit risk management problems. We treat the problem as a classification problem and use the data from LendingClub company. The six methods are the neural network, random forest, XGBoost, LGB, CatBoost, and Logistic Regression. The result of the experiment shows that the accuracy of the six methods all exceed 88%. Furthermore, the neural network gets the best AUC score, which can better meet the demand of commercial banks’ credit risk management.

This paper builds a novel deep learning model to test factor effectiveness as a basis for stock return prediction. The model uses LSTM modules to capture deeper connections between factors and stock returns other than only linear relationships and create a novel index for testing factor effectiveness. And based on the high effective factors extracted, a LSTM network is constructed to predict stork returns. The result is compared with the traditional method that depends on IC (Information Coefficient) to capture a strong linear relationship between variables. The model outperforms the traditional IC model both in the prediction accuracy and the reduced model training epochs. The ability to extract more effective information from data is an important implication for the business world, especially the stock forecast since insufficient data is often a large obstacle.

The unemployment rate reflects the utilization rate of labor resources in a country or a region, and is the most important indicator for predicting economic output. This study will use the urban unemployment rate from 1976 to 1986 to conduct an in-depth case study, try to explore the main factors affecting the unemployment rate in the United States, and classify the unemployment rate from 1976 to 1986 by region. The research mainly uses principal component analysis, K-nearest neighbors, support vector machines, decision trees and other methods, combined with some model evaluation methods, such as confusion matrix and Precision-Recall. The purpose is to find the most accurate model to classify and explain the United States from 1976 to Unemployment rate in 1986. In addition, this research will provide some useful suggestions for controlling the unemployment rate on the basis of in-depth analysis. In summary, the main factors are public capital and non-agricultural employment, both of which are negatively correlated with the unemployment rate. Except for the high unemployment rate in the United States in 1976, due to the steady recovery after the oil shock, the unemployment rate during that period was relatively stable.

Nowadays, the stock market generates a large amount of data every day, and the growth rate of these data is amazing. This situation leads to many problems faced by stock investors and companies, such as lack of ability to analyze massive data, fast data update, redundancy of data, and reduction of effective utilization of data. In order to make a better prediction, this paper constructs a bionic computing intelligent algorithm based on support vector machine, establishes an improved GASVM parameter optimization model based on the support vector machine model, and finally applies it to the financial market to predict the short-term stock price in the selected stock pool. The empirical results show that the prediction results are higher than the random prediction accuracy under different parameter optimization algorithms, which shows that it is effective to optimize the model by optimizing the parameters of support vector machine. Compared with neural network model or other traditional artificial intelligence algorithms, the model established in this paper is more economical and applicable, and can give investors guidance to a certain extent.

The outbreak of the Covid-19 caused the emergence of many related pandemic prevention policies, which evidently influenced the demand and supply sides of the agricultural products and may therefore make the relative products’ price fluctuate. This paper chooses the egg price in Henan province in China as a detailed situation to do the empirical research by using vector autoregression model (VAR), and try to find out the relationship between egg price volatility and the number of indigenous confirmed Covid-19 cases, Granger causality Wald test is also used in order to test the causal relationship between these two factors. We find that the fluctuation of egg price in Henan province was not positively related to the number of confirmed pandemic cases, but has the positive relationship with the supply of the eggs. Our findings suggest local governments should place much emphasis on ensuring stability of the supply side of the agricultural products in times of crisis.

Powerformer is a new type of generator developed by ABB. Start with the accuracy of Powerformer stator single-phase grounding protection, a new solution based on improved VMD and LightGBM is proposed. Firstly, synthetic kurtosis is introduced to adaptively optimize the decomposition level K, and the IMF component corresponding to the fault zerosequence current is obtained through VMD; Then, extract the IMF synthetic kurtosis and sensitive IMF comprehensive energy relative entropy and the comprehensive concave-convex coefficient respectively, which are used as three characteristics; Finally, the characteristic vector is formed and input into the LightGBM model to obtain the fault identification result of the Powerformers. Through the simulation in MATLAB, it shows that the accuracy is improved through this measure, which has better anti-noise performance than other schemes.

The COVID-19 pandemic has severely impacted our lives, and many industries are experiencing instability. We are curious to see whether retailing businesses, one of the most robust industries in US, have been affected, and if affected, how have the industry been affected. In order to gain the required insight, we first acquire stock prices of six representative retailing companies in US, then we apply ARIMA model on the data to forecast their trends in the near future, which will imply the general robustness of the industry. This procedure includes testing the stationary of time series data of stocks, and finding the suitable ARIMA parameters for each stock, using various methods. Accuracy metrics are brought into discussion to determine how accurate our forecast is. Finally, we draw the conclusion that ARIMA model, being a suitable method for our case of study, has given us desirable result: the stocks of 6 selected retailing companies will perform steadily with slight increase at the end of the year. There are several practical values or our research: By applying ARIMA models on stocks of retailing companies, we discovered that application of such models on retailing industry is not only pragmatic, but effective; and the results of our analysis provide future researchers with insight of economy of this era of pandemic.

The stock market is an important part of the financial market, and its fluctuation is closely related to the market economy. Predicting stock prices is an important part of many analysts and researchers' work. Choosing the right investment strategy is a challenging task. Therefore, this paper uses the historical trading data of the four largest market capitalization companies on the NASDAQ exchange as research objects to predict and analyze their short-term price trends. According to the trading data, the methods of linear regression, lasso regression, decision tree, XGBoost and RNN are selected, and the accuracy and feasibility of investment strategies are compared horizontally. After empirical research, it is concluded that the RNN-based neural network model has good prediction accuracy. The results of this paper can provide a reference for stock market strategy selection to a certain extent.

The theme of this paper was to examine different forecasting techniques to predict future stock returns based on past returns and numerical news indicators to construct a portfolio of multiple stocks in order to diversify the risk. This was done by applying supervised learning methods for stock price forecasting by interpreting the seemingly chaotic market data. Predicting the Impact of News articles on the Closed Price of the Apple Inc. Stocks using Naive Bayes Classifier. First of all the News Articles dataset and Historical Stocks Dataset were merged into a single dataset on the 'Date' column after making some necessary changes to them. Further, methods such as ensemble classifier which mainly entailed knowledge Nearest Neighbour(k-NN) were applied. Afterwards Natural Language Processing was applied on the News Headlines text and obtained a Bag of words containing 20000 most common words from by converting them to vectorized form. Then later the Naive Bayes model was trained (Gaussian, Multinomial or Bernoulli each in different files) by splitting the dataset, 80% as training dataset and 20% as test dataset. Overall, it was observed that the neural-based models achieved slightly better accuracy and macro-average F1 than the standard and ensemble models, and significantly better accuracy than the ZeroR baseline. Therefore, the news articles such as the model helps in classifying the news articles to a profit or a loss. This was done by calculating the difference in closed price of present day with the previous day.

In recent years, machine learning and data mining methods have been applied to stock price forecasting problems. However, due to the time series characteristics, data noise and multicollinearity of stock price data, many traditional machine learning methods cannot accurately predict the trend of stock price movement. This paper applies the Least Absolute Shrinkage and Selection Operator (LASSO) penalized logistic regression model and the ridge penalized regression model to the stock price prediction problem and evaluates their performance. First, we select the stock data of CMA, EQR and IRM from 2016-07-21 to 2020-07-13. Then we apply the TTR package to calculate 17 technical indicators. After that, we divide the data set into the training set and test set, train the model using the training set, and evaluate the performance of models on the test set. Finally, we introduce Support Vector Machine (SVM), Random Forest (RF), and Artificial Neural Network (ANN) as comparative models and evaluate the predictive performance of models by evaluating indicators such as sensitivity, specificity, and ROC curve. After the above work, we find that LASSO penalized logistic regression has the best prediction performance, reaching an accuracy of 0.81 on the EQR data set, and the ridge has a prediction performance of 0.79, which is much higher than RF, SVM and ANN.

Publisher's Note: This paper, originally published on 23 August 2022, was withdrawn on 17 January 2024 per author request.

Artificial intelligence (AI) appears more and more in people's life. Many academic articles point out that there are signal to noises in the financial market itself. The paper aims to find whether or not we can avoid the emergence of signal to noises by using different factors or models. We will select several machine learning models, including linear regression model, XGBoost in recent 5 years, and long short-term memory (LSTM) in time series. Using exponential moving average as a factor, this paper compares the performance of different machine learning models in the stock market. Finally, we find the results simulated by the regressor models are similar.

Based on big data technology, sort out the operation and parameter data of distributed photovoltaic in each district and county, and complete the analysis of influencing factors of distributed photovoltaic bearing capacity and thermal stability calculation and analysis in combination with the development background and current situation of regional distributed photovoltaic. Build the calculation algorithm and data model of distributed photovoltaic carrying capacity, mine the acceptable distributed photovoltaic capacity of each district and county, form the calculation boundary conditions of each district and county, and connect to the power grid orderly and reasonably. It ensures the safe and stable operation of the power grid and the full consumption of distributed photovoltaic, which is of great significance to guide all walks of life to actively participate in the orderly grid connection of distributed photovoltaic.

In addition to the body text, web pages also contain a lot of noise information such as advertisements and navigation bars. Accurately extracting text content from web pages is a key technology to improve the quality of web page analysis. The web page itself is a highly heterogeneous special text, and different types of web pages have different web page structures, which increases the difficulty of web page text extraction. After a lot of analysis, we found that there is a potential correlation between the body text and the tag path and text block density, so we propose a webpage text extraction method based on the tag path feature and the text block density feature. Combining the advantages and disadvantages of tag path features and text block density features, we design a fusion strategy to solve the problem of low accuracy of web page text extraction. The method does not require training, and improves the efficiency of webpage text extraction. The experimental results on the dataset constructed in this paper show that the classification accuracy of this method reaches 81.11%, the recall rate reaches 83.15%, and the average accuracy on all datasets is 17.7% higher than that of the BDF algorithm and 6.21% higher than that of the CEPF algorithm, the experiments show that the method has strong generalization ability.

The relationship between financial accounting and management accounting is inseparable in actual work, so we should seek mutual reference and application in enterprise management and services, and strive to achieve the dual needs of management and finance with the lowest human and financial cost. With the gradual convergence of the interests of the stakeholders of the enterprise and the operators of the enterprise, the blurring of the boundary between contemporary management accounting and financial accounting has become the basis for the integration of the two. At the data level, the original information sources of management accounting and financial accounting are basically the same; at the process level, management accounting and financial accounting communicate through various management accounting tools, enabling enterprises to make better use of accounting information to assist decision-making; at the reporting level, management The intersection of information generated by accounting and financial accounting constitutes the complete internal and external reporting of the enterprise. Data mining provides data support for the integration of management accounting and financial accounting. In this context, the deep integration of the two can effectively reduce data redundancy, promote the integration of corporate accounting processes, and restrict the distortion of accounting information, giving full play to the management function of accounting.

While numerous security products, including data leakage prevention, have been added to corporate cybersecurity strategies, securing confidential data and assets remains a major challenge for businesses and organizations. According to a survey by a research institute in the United States, most of the most costly cybercrime cases are caused by theft by insiders, followed by DDoS and Web-based attacks. In this paper, the LightGBM algorithm is used, and the feature extraction uses the bag of words and the IF-IDF model to construct a malicious operation behavior detection model. By training with the classic SEA training set, the results show that our model is more than 97% accurate. And compared to the popular classification models, our model has higher performance.

This paper sets up the research progress on building an intelligent power system with new energy as the main body to ensure national energy security. Under the scenario of "double high" and "double peak", it is the foundation for the stable operation of the new power system to promote the sinking of the production management center, strengthen the construction of the whole service intelligent core team. With artificial neural network, the paper based on company’s a grass-roots unit situation, designed the whole business intelligent core team quality and capacity analysis, and applied to the specific intelligent team construction, improve the quality of the work of a company team personnel at the grass-roots level and motivation, to enhance the system stability, and even the whole industry for the company intelligent core team building and stable sustainable development provides a powerful reference.

In this paper, we propose an automatic modulation recognition (AMR) algorithm based on multi-scale convolutional neural network cascaded bidirectional gated recurrent unit (MSCN-BiGRU) network. Firstly, the network extracts and fuse the multi-scale features of the raw signal I/Q data. and then obtain the global features of the fused features at a larger scale through the BiGRU network. Finally, the softmax classifier is used to complete the recognition. The experimental results show that the average recognition accuracy of the algorithm on the RML2018.01A dataset is 62.42%, surpassing CNN by 3.89% and ResNet by 2.8%, and the maximum recognition accuracy can reach 97.5%.

At present, most privacy protection models only consider static data, and it is easy for attackers to deduce some privacy data through data tables. In view of the changing characteristics of data, this paper proposes a set of dynamic data processing strategies, which are processed according to whether the data appears for the first time and the frequency of occurrence, and the KL divergence is adopted in the process of calculating similarity. Based on the dynamic data processing strategy, this paper proposes a (p,α)-Sensitive k anonymity algorithm for dynamic data. Through the analysis of the experimental results, it can be seen that the algorithm can protect the privacy of dynamic data and increase a certain amount of data-availability.

Like the existing software simulation experiment platform, the remote camera can also provide a realistic scene effect, giving the experiment user a sense of presence, thus greatly improving the quality and effectiveness of the experiment. It is verified and evaluated in the actual measurement environment that the research on the sampling method of the characteristic value samples is less involved. Thus, there is a certain gap between engineering practice and the application of instrument measurement. In order to solve the problems of less measurable nodes, less fault feature information and fault information from simulation in analog circuits, this paper chooses the low-frequency noise spectral density that can best reflect the fault feature information as the feature parameter through the study of circuit noise mechanism and noise superposition theorem. The paper also proposes an improved average period cross-spectrum method based on a low-frequency noise detection technology. With low-pass filter, low noise amplifier module, ADC sampling module and FPGA platform, the utilization rate of experimental facilities is high. Proved by experiments, the technology in this paper can reduce the number of facilities needed for the same experimental task, and reduce the cost of school laboratory construction and maintenance.

Singular spectrum analysis (SSA) is a nonlinear time series research method. Traditionally, the number of eigentriples to be reconstructed in the reconstruction step of the SSA method is usually decided by a given contribution rate threshold. However, this approach has inevitable subjectivity, and it will generate additional error. In order to remedy this deficiency, we propose an optimal number selection way by minimizing in-sample data error. We calculate the mean absolute percentage error (MAPE) values corresponding to all optional numbers of eigentriples using in-sample data. Then, the eigentriples corresponding to the lowest MAPE value are used as the reconstruction eigentriples. Meanwhile, the prediction coefficients corresponding to the lowest MAPE value are used as final coefficients in the linear recursive prediction formula. Based on this, out-of-sample data is predicted. The improved SSA method, the classical SSA method, and the ARIMA-GARCH model are compared in terms of forecast accuracy and computational time, using time series data from the closing prices of Shanghai copper futures. The results show that the improved method outperforms the classical ARIMA-GARCH and SSA models in forecast accuracy.

China has thousands of years of liquor culture, and almost 69 percent of Chinese alcohol consumption is liquor. At the same time, many liquor companies are listed companies, and the stock price is changing. So, stock forecasts for the liquor industry are very important. This article will predict the changes in stock prices in the liquor industry since the epidemic based on the long-short term memory (LSTM) model. First, the data of Moutai, Wuliangye, and Yanghe from June 2020 to June 2021 was selected as a reference. The return rate was calculated according to the close price, and the training set and test set were divided in a ratio of 4 to 1, and then the LSTM model can be constructed to make predictions after data normalization and convert it into the supervised problem. Finally, the results of the neural network prediction model show that the LSTM neural network method can well fit the fluctuation trend of the stock price of the companies, and the mean square error is small, which shows the prediction accuracy is relatively high, and it can reflect the fluctuation trend of stock price to a certain extent. So, this article can be a reference about the stock price for investors who preparing to invest in the liquor industry, mainly in leading liquor companies.

Face recognition has become an important component form of current biometric identification technology, the study of computer face recognition algorithms has been a hot topic of research and attention in academia and industry. However, the applicability of current face recognition algorithms to scenes still has much room for enhancement and improvement, especially in some special scenes, the accuracy of face recognition is poor, which is due to the influence of objective factors such as lighting and image angle on the one hand; on the other hand, interference caused by subjective factors such as facial movements and occlusion of the recognized person. To address the above problems, this paper designs a face recognition algorithm based on computer technology, which has significant application advantages at the index level of rejection rate, rejection times, comparison time and matching similarity, as well as the level of anti-interference.

A suitable pig house environment is an important way to ensure the health of pigs and improve the welfare and production efficiency of pigs. Intelligent pig house environment control technology and method is the only way for the pig industry to transform traditional breeding methods and realize refined breeding. According to the relationship among the air temperature and humidity, ammonia concentration, carbon dioxide concentration, particulate matter concentration and pig growth performance and health in the pig house, this paper establishes the air quality safety control point of the pig house, and realizes the intelligent control system of the air quality of the farm. Through the deployment and control of intelligent sensor monitoring points, the full coverage of the pig house environment is achieved. Pig farm managers can log in remotely through 4G mobile phones to obtain real-time environmental information of the pig house. The pig farm managers can automatically control the environment monitoring equipment through mobile phones or let the equipment automatically inspect the pig farm environment, and finally realize the remote intelligent management of the pig farm environment.

With the rapid development of information technology, ICO, as a typical token, has been one of the focuses of domestic and foreign supervision. This research focuses on the ICO Economic Crime Model which is based on the BP Neural Network. Through the study and analysis of 150 ICO examples, 13 evaluation indexes of the ICO economic crime model are selected, and the ICO economic crime model based on the BP neural network is established. The model uses the function approximation ability of the BP neural network to simulate the nonlinear function relationship between the evaluation index and the ICO legitimacy to realize the evaluation of the ICO legitimacy. The empirical results show that the BP neural network has high accuracy in diagnosing the legitimacy of the ICO, which can be adopted as a new idea to evaluate the legitimacy of the ICO.

The causes of track irregularity are determined by many factors. It is very difficult to evaluate and quantitatively analyze the causes of the natural environment. On this basis, the track irregularity is regarded as a grey system, and the GM (1,1) model is used to model and evaluate it. However, the traditional grey model with equal time intervals is used to evaluate the deterioration trend of track irregularity, which can not fully reflect the changing trend of track irregularity in a single maintenance cycle. Therefore, to reduce the impact of various factors and improve the accuracy of the assessment results, this paper combines the actual characteristics of the trend assessment of track irregularity based on the grey model theory, which optimizes and improves the traditional grey model from four directions: time interval, calculation of background value, determination of initial value, and weight of detection data. The evaluation accuracy of the model in this paper is more than 1. 64% higher than that of other comparison methods, which can reduce the solution error of the model, improve the sensitivity of the model, and make the model accurately and timely predict the deterioration of track irregularity.

With the development of social economy, people's concept of investment and financial management has gradually strengthened. Machine learning can be applied in the volatility. In our paper, we use hybrid model based on XGBoost, LightGBM and TabNet for predicting realized volatility. We conclude some related works and represent our hybrid model in the detail. In the experiment process, we compare our hybrid model with other models, and the result shows that our hybrid model owns the lowest RMSPE score 0.198 among all compared models.

Understanding the interaction to the context and the co-occurrence of actions is necessary for human action detection. This paper proposes a Context-Action Dependency (CAD) model for spatio-temporal action localization tasks. It leverages context features and action relations through two attention-based modules. A context-dependency module learns the relation between actors and attentive context information. An action-dependency module models each action class separately and explores the relations within co-occurring actions. This paper introduces an action detection framework that integrates the CAD model to improve the precision of action localization. To verify the effectiveness of our solutions, we ran experiments on two datasets AVA and UCF101-24. Our experiment shows that CAD has better performance than the baselines.

This research mainly focuses on comparing the performance of the Black-Scholes Model and the Heston Model when pricing options. In practice, the researchers used a trust-region algorithm to numerically calculate the parameters required in both models. With the parameters, the researchers used them to hedge an imaginary option attached to AAPL and used the hedging performance as a measurement. Also, percentage error was used to measure the performance of the two models. The researchers reached that the B-S model behaves better in short term and the Heston model performs better in long term. By comparing the advantages and disadvantages of these two models, we are allowed to give insightful suggestions on improving the models to give more precise results.

Breast cancer has become the most cancerous cancer in the world. Along with the increase in the number of confirmed breast cancer patients, the early diagnosis of breast cancer is extremely important. Machine learning methods have a wide range of applications in breast cancer tumor diagnosis, of which artificial neural networks are an effective tool for computer-assisted breast cancer diagnosis, and error backpropagation (BP) neural networks are the most widely used models. This paper provides a diagnostic decision based on the combination of rough set (RS) and BP neural network, using the rough set genetic reduction algorithm as a feature selection tool, and using the minimum number of features and the number of identifiable features in the optimal distinction matrix as the fitness function to achieve feature reduction, remove redundant attributes, further improve the diagnostic accuracy of breast cancer on the basis of a single neural network classification model, and select the five feature factors with the greatest influence on breast cancer. Experiments with breast cancer datasets in the UCI database show that this method effectively improves the accuracy of classified diagnosis compared with the original BP neural network.

As one of the most popular classification algorithms, random forest has been widely used in various fields of life. It not only has a relatively perfect theoretical background, but also has a good performance in data classification. However, the classification and prediction of unbalanced data sets is still difficult. In order to solve this problem, a classification and prediction model of unbalanced data sets based on adasyn-rf is proposed. Firstly, adasyn algorithm is used to deal with unbalanced data sets, and then RF model prediction is constructed. In order to illustrate the superiority of the model, comparative experiments are added. That is, the original data set and the data set processed by smote, boundary smote and adasyn are combined with two prediction models: random forest model and xgboost model. The experimental results show that the classification and prediction effect of unbalanced data sets based on adasyn-rf model is the best.

The research is conducted during the rage of COVID-19 throughout the world. The world meets new challenges from COVID-19 from every dimension, especially the economical world. In the economic world, the most related part for the influence that springs from COVID-19 is the stocks belonging to the healthcare sector. Aiming at doing the return prediction for healthcare sector stocks, the study chooses Long Short-Term Memory (LSTM) Algorithm to introduce machines to adapt the pattern and make predictions. The study selects 6 less volatile while keeping high average trading volume stocks from the healthcare sector. Using the LSTM learning model to learn the past 5 years’ data and make the prediction to the future 5 days. The data consist of 65% of the company's data from five years ago as the training set, and the last 35% of the data as the test set. The study compares the actual data to the predicted data and sees the error by calculating root mean square error (RMSE). The result draws the conclusion that the model will perform more precise prediction when the picked stock has a clear price trend and less fluctuation. The application for this study is to provide a short-term trading strategy and manage the risk for short-term stock investment by using the LSTM model.

User interest model is the core of personalized information service, and its quality has a direct impact on the effect of personalized information service. However, the traditional interest model often only considers the user's interest in the text and ignores the influence of contextual information. Therefore, in order to avoid the interaction between context and interest features, this paper proposes a multi-level vector space interest model representation method. Then, the user's interest is weighted from many aspects by using the user's contextual information. Finally, the personalized retrieval experiment is carried out by using the constructed interest model. The experimental results show that the retrieval precision based on the model reaches 83.72%, which proves the validity and reliability of the model.

To express the right information about enterprise data and trace the changes of data right in the whole life cycle of data, this paper designed a data right provenance model based on the W3C PROV model. The PROV-R model defined the core class to support the representation of data right with a new set of relations to express the interaction between the classes. In order to trace the complex right relations, the model uses some inference rules to add the new relational statements. Finally, a data trading process shows the application of this data right provenance model.

With the rapid development of the unmanned aerial vehicle (UAV) industry and artificial intelligence industry, UAV has been used in the civil field. For civil UAVs, the distribution of logistics UAV is a hot research topic. Based on the UAV data set, the value function model of logistics UAV task planning is established by deep learning. It is found that there are obvious differences in value function for different deep learning methods. Moreover, it is also found that using the Random Forest Model has a good effect on the modeling of the value function. On this basis, the task planning model of logistics UAVs based on K-Means is established. In order to find an optimization algorithm suitable for the logistics system, this research uses an improved Genetic Algorithm, improved TS Algorithm, and Particle Swarm Optimization Algorithm to solve the model. By comparing three intelligent optimization algorithms, it is found that the improved TS Algorithm not only converges faster for this model but also has fast convergence for a large number of logistics task points by comparing the flight route and convergence curve of the objective function, which can meet the needs of largescale tasks in the actual logistics system.

There are many types of explosion-proof equipment, so the problems in use are not the same. In view of these situations, this project takes the explosion-proof electrical equipment as the research object, and develops a set of integrity evaluation system of explosion-proof electrical equipment. This paper presents a defect detection method applied in the integrity evaluation system of explosion-proof electrical equipment. The proposed method first designs a lightweight network capable of achieving high-definition image detection under the YOLOV4 detection framework. Then, according to the specific data set, a priori box corresponding to the data distribution is generated. Finally, the distance-based K-means clustering method is used to extract the anchor block. The STN-OCR detects and identifies the defects from the natural images, thus applying the defect detection method in the explosion-proof electrical integrity evaluation system. This method improves the traditional evaluation process of traditional explosion-proof electrical equipment, which is cumbersome and inefficient, and has a certain impact on the application in some other fields in the future.

Sustainable production and consumption (SPC) is both a policy area and a multi-interdisciplinary research agenda that seeks to address the transformation of society toward economic and social sustainability. The SPC field encompasses diverse research topics and theoretical approaches, often characterized as falling into stronger and weaker versions, defined by a commitment to more or less systemic economic change. This study researched the economic impact of big data technology on sustainable production and consumption in live streaming E-commerce consumer behaviors. Based further on stimulus-organism-response (S-O-R) theory, the live streaming E-commerce’s anchor-controlled factors and viewers’ engagement provide the stimuli for the cognitive appraisals of the viewers (potential customers). Two cognitive appraisal results are trust and perceived value, known as an organism in the context of S-O-R theory, which, together with the stimuli, influence the flow experience behavior of viewers (potential customers). Flow experience is shown supportive of explaining impulsive purchasing. Towards this end, this study extends the S-O-R structure into S-O-B-C (stimulus-organism-behavior-consequence) paradigm. While organism and flow experience behavior (O and B) are domains of viewers and customers, stimuli (anchor’s reputation and professionalism, anchored interactivity, and the displays of offers) and consequences (impulsive purchasing and fan-becoming) are domains of the live streaming ecommerce. The findings resulted from a convenience-based sampling of customers and viewers who frequented popular live-streaming apps in China: Tao Bao, Jing Dong, Moquije, Little Red Book, and Jumeiyoupin. The data were subjected to neural network (NN) simulation and structural equation modeling (SEM), yielding similar results. SEM statistics and path configurations provide model fitting supports and mediating phenomena involving trust, perceived customer value, and flow experiences. Results confirm the significant values of cognitive appraisal theory, flow theory, S-O-R theory, and the extended S-O-B-C paradigm.

The stock market’s prediction and portfolio construction afterward have been seen as important and effective since Markowitz’s portfolio theory. While many models of prediction and portfolio combination exist, the most appropriate prediction models for specific firms are not definite. In this paper, the stock price of six leading enterprises in different fields were predicted by four statistical models including Multiple linear regression (MLR), K-Nearest Neighbor (kNN), Auto-Regressive Integrated Moving Averages (Auto-ARIMA) and Long Short Term Memory Networks (LSTMs). Crossvalidation was conducted using model assessment statistics R2, RMSE and sMAPE to compare the prediction performance of each model. The result revealed that LSTM performs the best among the four models on the six selected firms. The expected return generated by LSTM was then used to construct the portfolio model by allocating risks and returns. The final portfolio has shown the detailed distribution of each firm’s stock. Researchers can do further refinements by adding more models and firms based on the methodology in this paper.

Great Wall Motor is selected as the research object, the entropy method is used to determine the weight of each financial indicator of the company, then the final financial indicators were screened by correlation analysis, and the indicator weights were determined again; the efficacy coefficient method was adopted to determine the standard values of each indicator, determine the efficacy coefficients of each indicator, and then calculate the scoring values of individual indicators, and finally, the comprehensive scores of the research object were summarized. Through calculation, the financial risk warning level of Great Wall Motors is a slight risk, indicating that financial risks may occur at present. Managers should pay attention to some abnormal financial indicators and take targeted measures for them to resolve financial risks in a timely manner.

This paper examines and analyses the volatility of precious metal assets and the portfolios management. The innovative study of precious metal assets and equity portfolios is instructive and forward-looking under the background of global COVID-19. We use GARCH model, Markowitz model, CAPM Model and Fama-French Three-Factor Model to study volatility, asset forecasts, and the relationship between asset prices and markets. The results show that, firstly, precious metals are less volatile than stocks; secondly, precious metals, especially gold, make up a large percentage of rational riskhedged portfolios; and finally, the stock prices have an almost linear relationship with the market, but precious metal prices do not have a linear relationship with the market. Our conclusion suggests that investors should consider precious metal assets, especially gold, in addition to stocks when considering their portfolios, which helps to hedge investment risk and provides some guidance to the market.

Based on the actual statistical WIM data, this paper studies the transportation profit, transportation cost and road maintenance cost under different vehicle axle limits and analyses the comprehensive economic benefits. The formula of transport profit TP, transport cost y and road maintenance cost 𝑅𝑀𝐶 is put forward to calculate the three parts of cost, and the relationship and trend of the three parts of cost and comprehensive economic benefit are analyzed. The results show that the transportation profit increases logarithmically under different axial limits, the transportation cost increases like a linear function, and the road maintenance cost increases exponentially. Reasonable load limit value can not only prolong road life, reduce maintenance cost, but also maximize comprehensive economic benefit. Finally, the axial limit value under optimal economic benefit is proposed to provide reference for related departments.

China's rural informatization construction has a good foundation. However, China's rural informatization construction still has a long way to go. It is not only facing major development trends such as integration, specialization, networking, multimedia, practicality, popularization, integration and whole process, but also facing the urgency of farmers' increasing agricultural income and farmers' demand for information services to open up a larger consumer market. Moreover, increasing grain production and farmers' income are not only the focus of the "three rural" issues, but also related to China's steady economic development and social stability. In the selection of variables, this paper uses the GDP of the primary industry to quantify the income of farmers, and uses the number of broadband households to quantify the degree of rural information networking. In terms of the development trend of rural information construction, this paper will study the rural information networking, test it by using Vif multicollinearity expansion factor method, and study the relationship between farmers' income and the degree of rural information networking by ridge regression method.

Due to the sudden outbreak of COVID-19, there is a high volatility in stock price of vaccine manufacturers in China (Between December 15, 2020 and December 13, 2021, average monthly volatility of these companies is 986). The aim of this paper is to compare the price prediction result of four algorithms: Multivariable Regression Model (MLR), Auto Regressive Integrated Moving Average Model (ARIMA), Back Propagation Neural Network Model (BP-NN), Random Forest Regression (RF), and decide which one has a better performance. Data from December 2020 to December 2021 is collected from Wind and the 8 stocks is selected in leading companies in vaccine industry. It turns out that BP-NN Model gives the best result in predicting stock price of vaccine manufacturers, measured using commonly used indicator, i.e., root-mean-square error (RMSE) and R Square (R²). So next time in the similar situation, BP-NN can be seen as a powerful tool to help us make decision. This paper would help investors build an optimal strategy in selecting stocks in terms of pharmaceutical industry.

With the introduction of China's "One Belt, One Road" policy, the increasingly close economic interaction may enhance the linkage effect of the stock markets of the countries along the route. Exploring the interaction and transmission direction of stock market volatility in countries along the route can reflect the macroeconomic correlation of countries along the route and provide more effective suggestions for stock market development. This paper selects the CSI 500 Index, the Indian SENSEX Index and the Russian RTS Index, and mainly uses the BEKK-GARCH model and wavelet analysis to find that the changes in the Russian stock market are the Granger reasons for the changes in the Chinese and Russian stock markets. The wavelet spillover test shows that the Chinese, Indian and Russian stock markets all show volatility aggregation and persistent spillover effects. The two-way spillover effects between China-India stock market, China-Russia stock market, and India-Russia stock market are still significant on short-, medium- and long-term time scales.

With China 's financial market development, many investors look into the high-yield stock market, so it is important to find a more accurate way to predict stock prices. This paper takes the Shanghai Composite Index as the research object, selects the data from November 5, 2020, to November 5, 2021, and randomly selects three groups of continuous 15 - day data for research. Four models, namely, GM (1,1), RBF, GM (1,1) -RBF based on equal weight and GM (1,1) -RBF based on composite weighting method, are adopted to fit and predict the opening price. Compared with GM (1,1), RBF and equal weight GM (1,1) -RBF, the optimization model based on the composite weighting method has improved the prediction accuracy and accuracy, with an average accuracy increase of 3.7 % and a strong generalization of the model. The innovation of this paper is to combine the stable but large error algorithm with an unstable but small error algorithm, which has higher robustness than the single model prediction and use this method to achieve the accurate short-term prediction of the Shanghai Composite Index. The improved financial stock price combination forecasting method proposed in this paper can provide a reference model for financial stock price prediction and provide some suggestions for investors ' quantitative investment.

China Stock Exchange was established in the 1990s and developed rapidly in the past 20 years. It is an important place for domestic equity financing and investment by investors, however, China's stock markets were too short to adapt to CAPM model. In this paper, Whether China's stock market is suitable for CAPM model is continuously analyzed and studied. We selected four representative stocks in Shanghai and Shenzhen 300 from 2010 to 2021 to confirm the applicability of CAPM model. The results show that only one of the four stocks conform to CAPM model theory, and the return of the other three stocks does not have a high degree of compensation for system risk. This does not conform to the CAPM model, so we conclude that the CAPM model is not applicable to CSI300, and the systemic risk accounts for a large proportion of the total risk. We draw our conclusion that the CAPM model theory is not yet applicable to both stock markets in China. In addition, according to the research results, this paper also summarizes the problems in the process, in order to find the most appropriate theoretical model for China's A-shares.

The objective of this study was to investigate whether the estimated options price from the deep learning, method fits with the actual price based on high accuracy and determine the feasibility of applying deep learning in practical options pricing. We predicted the price of China A-shares with the long short-term memory method, which is the recurrent neural network in deep learning, using the past 6 years call and put options prices of China A-shares. The results from the long short-term memory method illustrate that the method could predict the prices of China A-shares with high precision and we could reveal that the deep learning methods are feasible in options pricing. Furthermore, we conclude that the analysis of China A-shares pricing and long short-term methods has provided a deeper insight into promoting the deep learning methods in the actual options market. In addition, future research should provide more details about applying the other deep learning methods for the other options with an appropriate range of historical prices to have stronger support about the deep learning methods are reliable in the practical options market.

With the effective control of the new crown epidemic, how to develop China's financial industry in the post-epidemic era still lacks a suitable regional index system. Taking Beijing, where the financial economy occupies an important position, as an example, this paper selects 12 typical relevant indicators from the four dimensions of structure, scale, vitality and efficiency, and constructs a model based on Radial Basis Function (RBF) and Support Vector Machine (SVM). The threelevel Bagging integrated learning area financial development index system. First, the coefficient of variation method (CVM) is used to determine the weight of each index to measure the level of financial development, and then based on RBF and SVM, the Bagging integrated learning prediction is performed, and the financial development index is obtained. The results show that the RBF-SVM integrated model is more fitting than the single RBF or SVM. The financial development index of Beijing has generally shown positive growth in the past 20 years, and the financial industry has a good development trend. In this paper, we establish a financial development index system to judge regional financial development differences and predict the development trend of the financial industry, thus providing suggestions for the healthy development of my country's financial industry.

The rise of new technology is reshaping corporate governance regimes and firm financial behavior. Investment activities are the main driving force for the rapid development of the nation and the basis for the company's success in today's competitive market. This paper adopts Richardson's expected investment model and the pooled regression model, based on the data A-share listed companies from 2008 to 2020, with executive incentives as the main entry point, to empirically study the impact of the internal compensation gap of executives on investment in China. The results show that the internal compensation gap of the senior management team is negatively related to investment efficiency. The larger the internal compensation gap, the higher level of inefficient investment; In state-owned companies, the negative effect of the executive internal compensation gap on investment efficiency is more significant, and inefficient investment is reflected as over-investment. This paper provides targeted practical advice for different companies on how to improve investment efficiency through the system design of executive compensation.

With the rapid development of digital information technologies such as big data, cloud computing and artificial intelligence, quantum information technology and digital economy are profoundly changing our way of production and life. New forms of digital economy, such as industrial Internet, intelligent Internet of things and online shopping, have become new driving forces for high-quality economic development. As a branch of digital economy, Internet consumer finance is in the trend of continuous development. This paper mainly studies the influence of Internet consumer finance on residents' consumption through VAR model. VAR model constructs the model by combining every endogenous variable in the system with its hysteresis value and exogenous variable, which has a good prediction effect on interrelated time series variables. In addition, the dynamic impact of different types of random error terms on system variables can also be analyzed. Therefore, VAR model is used in this paper to explore the benign interaction between the development of Internet consumer finance and consumer consumption. Through a series of empirical tests and analyses such as ADF test, VAR parameter estimation and impulse response function, the linear relationship between Internet consumer finance and consumption level is quantified and the future development is predicted. A series of countermeasures and suggestions are put forward for reference by relevant researchers.