Simulation of regional financial behavior patterns based on recommendation algorithm and machine learning

doi:10.21203/rs.3.rs-2751098/v1

Download PDF

Research Article

Simulation of regional financial behavior patterns based on recommendation algorithm and machine learning

https://doi.org/10.21203/rs.3.rs-2751098/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

The rapid growth of regional economy can effectively solve the financial difficulties of some large enterprises, and can also provide important guarantees for the sustainable economic development of some regions. In modern financial regional entities, commercial banks, conglomerates, and third-party technology companies are the main modules. Financial development issues in different regions need to be addressed through the design of solutions. In the cross domain and even cross border activities of regional economy, it is necessary to integrate some existing financial tools, as well as financial services such as big data and digital technology, to integrate. In the construction of a new regional financial model, regional financial health is a major development issue, and it is necessary to take dredging regional financial development as an important field to promote the rapid development of the real economy. In recent years, learning in natural language processing has achieved cross era development. With the growth of the complexity and dynamics of network information, computer algorithms have become an important key issue in solving network information. Although high-quality recommendation algorithms have been widely used, in order to solve problems that traditional algorithms cannot solve, this paper proposes a machine learning based algorithm based on the needs of financial entities. At the same time, in order to enable the rapid development of finance, there is a new development model that conducts behavioral simulation calculations based on economic development in order to achieve coordinated economic development.

Recommendation algorithm

Machine learning

Regional finance

regional economies

In recent years, consumer demand has been increasing, and consumers' personal attitudes have also been constantly changing. With the rapid development of the economy, major banks have undergone significant changes in consumer financial products based on user needs. The financial business has gradually entered people's daily lives, and various loan products launched by credit institutions have caught people's eyes [1]. With the advent of the 21st century, more and more people have made loans. This loan model is not linked to everyone's poverty and wealth, but rather a new type of consumption model. Recently, with the explosive growth of credit business, the financial consumer services on the Internet are not only an important business for banks, but also a business for many financial institutions to survive and develop [2–3]. In terms of credit management, it is necessary to evaluate and investigate the borrower's repayment ability and willingness to repay, with the aim of ensuring that the repayment period and amount of the loan can be carried out on time. In terms of credit risk assessment, it is necessary to assess the loan applications of defaulting users based on their willingness to repay. In terms of credit institutions, Internet finance needs to be synchronized with regulatory measures and technologies, and there are certainly regulatory blind spots [4]. If the threshold for the credit industry is low and the quality of the people involved is uneven, then this risk cannot be accurately assessed. Customer managers who need banks must find an effective solution and investigate some defaulting users [5]. Therefore, the risk monitoring of lenders has become a very important and increasingly serious problem faced by financial and credit institutions.

In recent years, the phenomenon of data explosion is increasing day by day, but with the increase in the number of information, sensor technology and computer technology are rapidly developing, and the problem of excessive information becomes more difficult to solve. In order to solve the problem of information overload, artificial intelligence applies abstract technology to the issue of information transmission from the perspective of big data acquisition [6]. In order to analyze the historical behavior of users, it is necessary to model the products that users are interested in, so as to facilitate personalized comparison in the later stage. In order to combine the main functions of the algorithm with the interests of users, it is also necessary to launch unknown products recommended by users with similar products. At the same time, in the field of machine learning, data mining has become a new hot field with the rapid development of technology [7]. In the field of language recognition processing, although certain achievements have been achieved in these fields in recent years, it is still necessary to study the structural complexity of the Internet through the construction of training targets for labeling, as well as centralized learning of data [8]. If unmarked training data is used as a focus for recommendation algorithms, it is necessary to better control regional financial behavior, In order to facilitate the healthy development of finance.

The literature has evaluated the risk control risk of the financial and credit industry by proposing an evaluation model based on discriminant analysis, which leads to the concept of risk control. In order to enable this model analysis method to discriminate and predict, a multivariate model analysis method was conducted based on the operation of the enterprise in the literature [9]. Of course, in order for traditional discrimination methods to meet certain requirements, it is also necessary to make certain assumptions about most of the data in real situations. If some data cannot predict the collinearity relationship of variables based on application conditions, then the prediction ability of the data will become low. The literature compares the mortgage and pledge modes of accounts receivable, and expounds that there are certain clear concepts in regional finance [10]. For the concept that regional finance has a relatively long exploration stage, it is necessary to make the concept of regional finance clear and understood by most people in order to achieve the substantive content and financing behavior of regional finance [11]. According to the literature, in order to reduce the probability of risk occurrence in the supply chain, it is necessary to evaluate the financing model of regional finance and the reasons for its occurrence between enterprises and banks to avoid the phenomenon of information asymmetry. The literature uses filtering algorithms on a scoring matrix to recommend Bayesian learning classifications, text information, and important content [12]. If the final scoring performance of the learned movie text information obtained from the harmonic average is relatively high, it is necessary to compare the separately recommended models. In order for the calculated user to achieve the effect of weight fusion in the final result, it is necessary to process the filter method with the value of the cut and cold start. If the content to be accessed involves content such as sequences, history, logs, etc., it is necessary to further advance the filtering method and calculation with mutual speed [13]. The neural network model proposed in the literature can combine recommended systems with machine learning to facilitate the evolution of recommended systems in various directions [14]. Changes in the complexity of neural networks used in the literature can be combined based on the recommended features of deep networks to facilitate the scoring of user word encoders and item vectors, enabling the structure to be quickly trained and deployed. In order to change the operational form of matrices in the literature, it is necessary to decompose and cross exchange matrices. Use the model and expression capabilities of neural network elements to splice and introduce them [15]. If the decomposition model of the matrix and the model of the outer product operation are embedded and spliced, it is also necessary to create and combine the expression capabilities of the simplified operation [16]. Part of the model is enhanced through manual screening or the realization of memory capabilities through new features, in order to facilitate the continuation of ideas in subsequent models. In the early stages of machine learning, if a neural network model is used to aggregate data based on the learning model, it will surpass the previous model effect [17]. It can also be seen that any task can achieve results based on the application of the network.

3.1 Theoretical basis of recommendation algorithm

The cosine can be expressed based on the calculated characteristic quantity of the user. The similarity is calculated by measuring the cosine value of the angle between different vectors to obtain their similarity. The range of values of the angle between i and j calculated by the user will be close to the similarity of the cosine and the preference results. For spatial points between vectors, the absolute distance is expressed as:

$$\text{d}(\text{i},\text{j})=\sqrt{\sum {({\text{i}}_{\text{k}}-{\text{j}}_{\text{k}})}^{2}\in [0,+{\infty })}$$

Convert to the intermediate vector C:

$$\text{C}=\text{F}({\text{x}}_{1},{\text{x}}_{2},{\text{x}}_{3},{\text{x}}_{4}),$$

Calculating the hidden coefficients of the encoder and decoder will be weighted and summed based on the weight coefficients, resulting in a sequence output encoding vector that is:

$${\text{C}}_{\text{i}}=\sum _{\text{j}}{{\alpha }}_{\text{i}\text{j}}{\text{h}}_{\text{j}}^{\text{i}\text{n}};$$

Calculate attention weight:

$$\text{W}\text{e}\text{i}\text{g}\text{h}\text{t}:{{\alpha }}_{:\text{j}}=\text{S}\text{o}\text{f}\text{t}\text{m}\text{a}\text{x}\left({\text{K}}^{\text{T}}{\text{q}}_{:\text{j}}\right)$$

The transition state of random variables will be cumulatively increased in the next training state based on the expectation of receiving rewards. Since the transition state is random, Rs will obtain a corresponding weight accumulation coefficient, and the reward time will also decay over time:

$${\text{G}}_{\text{t}}=\sum _{\text{k}=0}^{{\infty }}{{\gamma }}^{\text{k}}{\text{R}}_{\text{t}+\text{k}+1}$$

The features of the matrix will be combined according to the decomposition recommended algorithm. If the basic idea of the algorithm is decomposed into two low-dimensional feature matrices, a fitting matrix will be formed. In order to complete the missing data in the matrix, it is necessary to predict the unknown data in order to achieve the algorithm prediction recommended for decomposition of the unknown matrix. In order to establish an objective function, the commonly used methods are to decompose the matrix and use the gradient descent method. The objective function formula is as follows:

$$\text{L}(\text{U},\text{V})=\frac{1}{2}\sum _{\text{i}=1}^{\text{m}}\sum _{\text{j}=1}^{\text{n}}{\text{I}}_{\text{i}\text{j}}{({\text{R}}_{\text{i}\text{j}}-{\text{U}}_{\text{i}}{\text{V}}_{\text{j}}^{⊺})}^{2}+\frac{{\beta }}{2}({‖{\text{U}}_{\text{i}}‖}^{2}+{‖{\text{V}}_{\text{j}}‖}^{2})$$

The triangular structure in a mesh can be summarized as a three node structure. Because a triangle is composed of three nodes, this local structure is shown in Fig. 1:

In this article, if the symmetric matrix of the user is a similar matrix, it is necessary to combine the structure of the similar network with the structure described in this article to obtain the calculation formula for the adjacent matrix. In the matrix, if the minimum value of the initial value is 0, it indicates that the initial value is an unknown programming function. The formula is as follows:

$$\text{g}\left(\text{x}\right)=\frac{\text{x}-\text{m}\text{i}\text{n}}{\text{m}\text{a}\text{x}-\text{m}\text{i}\text{n}}$$

To calculate the similarity between users, a cosine similarity matrix can be obtained.

$${\text{T}}_{\text{i}\text{k}}=\frac{{\text{P}}_{\text{i}}\bullet {\text{P}}_{\text{k}}}{\left|{\text{P}}_{\text{i}}\right|\times \left|{\text{P}}_{\text{k}}\right|}$$

In order to achieve a divided distance of two sample points, the SVM idea classifies them based on the planar distance reached by the sum of the distances between the two heterogeneous sample points. The heterogeneous sample points used as support vectors will become interval vectors based on the sum of the planar distances reached by the sample points, and the formula is as follows:

$$\text{r}=\frac{2}{‖\text{w}‖}$$

Using option speed as the degree of control for a certain stage allows some users with motif structures to more integrate higher-order control parameters on a similar basis. The formula is as follows:

$$\text{S}=\text{T}+{\alpha }\bullet \text{g}\left(\text{D}\right)(\forall \text{i},\text{k}\in \text{F},\text{i}\text{f}{\text{S}}_{\text{i}\text{k}}>1,\text{t}\text{h}\text{a}\text{t}{\text{S}}_{\text{i}\text{k}}=1)$$

When the user's predicted value reaches a certain threshold, compare the settings of i and j values, and if it is greater than the set value, recommend it to the user:

$${\widehat{\text{R}}}_{\text{i}\text{j}}={\text{U}}_{\text{i}}{\text{V}}_{\text{j}}^{⊺}$$

The algorithm in this article is compared with the recommended algorithm.

Table 1

Recommended Performance Comparison
Algorithm	BiasSVD	MF	SVD++	JMF-UCP	Algorithm in this paper
RMSE	0.3608	0.3477	0.3481	0.3495	0.2947
MAE	0.2374	0.2354	0.2314	0.2266	0.1930

3.2 Regional financial market prediction based on machine learning

SL is a model for training sample data using tags. In the model training algorithm, in order to predict tag data, review and classification can be conducted based on supervised learning by a mentor. The process framework of SL is shown in Fig. 2:

As the input of each level is a feature obtained at that level, an unsupervised learning mode has become a typical representation. Different letters represent different output meanings, and the loss function is expressed as:

$$\text{L}({\text{y}}_{\text{i}},{\widehat{y}}_{i}({x}_{i}\left)\right)=\left|{y}_{i}-{\widehat{y}}_{i}\left({x}_{i}\right)\right|$$

In the described model structure, in order to avoid overlapping phenomena of certain models, the calculation formula can be expressed as follows:

$${J\left(\theta \right)}_{min}=\frac{1}{N}\sum _{i=1}^{N}\lambda L({y}_{i},{\widehat{y}}_{i}({\theta }_{i}\left)\right)$$

When performing speed searches at different locations, comparing the dimensions of the subgroup with the weight coefficients of the neural network model, it can be concluded that the introduced value of the objective function is:

$${f(x}_{i})=\frac{1}{l}\sum _{i=1}^{l}{y}_{i}$$

In terms of risk management in the financial market, if the predicted hit rate has a certain significance, it is necessary to include the prediction in this article as an effective standard value. In addition, this article also calculates unequal coefficients based on the prediction criteria, and the calculation formula is as follows:

$$\text{H}\text{R}=\frac{100}{\text{N}}\sum _{\text{i}=1}^{\text{n}}{\text{d}}_{\text{i}}({\text{X}}_{\text{i}},{\text{Y}}_{\text{i}})$$

In order to predict the marketization trend of the experimental objects adopted in this article, it is necessary to conduct a dynamic analysis with my prediction results. The prediction results of regional financial market trends are shown in Fig. 3. If the short-term financial market is not consistent with the prediction results of the method in this article, it is necessary to compare the prediction results, as shown in Fig. 4:

When weighting and summing the data input by neurons, it is necessary to activate the input function and judge the output. If the given function value is x, the output result formula is:

$$\text{S}=\sum _{\text{i}}{\text{w}}_{\text{i}}{\text{x}}_{\text{i}}+\text{b}$$

The Sigmoid function, also known as the Logistics function, is defined as follows.

$$\text{S}\text{i}\text{g}\text{m}\text{o}\text{i}\text{d}\left(\text{x}\right)=\frac{1}{1+\text{e}\text{x}\text{p}(-\text{x})}$$

Suppress all values that are less than zero, if the principle of inhibition comes from neurobiology. When comparing the input and output values of a function, a phenomenon of gradient dispersion usually occurs, resulting in slow updating of the active function and a halt in training. In order to make the characteristics of functions widely used, the definitions are as follows:

$$\text{L}\text{e}\text{a}\text{k}\text{y}\text{R}\text{e}\text{L}\text{U}\left(\text{x}\right)=\left\{\begin{array}{c}x,x\ge 0\\ px,x<0\end{array}\right.$$

The interest characteristics of each modeling user can be represented by the following formula.

$${\text{p}}_{\text{j}}={\sum }_{\text{k}=1}^{\text{t}}{{\alpha }}_{\text{j},\text{k}}{\text{v}}_{\text{j},\text{k}}$$

On each predicted interaction node, the prediction module will calculate different proximity algorithms, and different similarities have different nodes.

Based on the ranking of the average and the recall rate, the model performance of UIPST can be obtained by comparing users' products with each other within the prediction time, as shown in Table 2:

Table 2

Experimental Results
Method		LSTM	Time-LSTM	CTDNE	STAMP	UIPST
Data-1week	MRR@10	0.344	0.398	0.411	0.633	0.741
Data-1week	R@10	0.555	0.566	0.512	0.761	0.854
Data-1mouth	MRR@10	0.369	0.251	0.444	0.521	0.751
Data-1mouth	R@10	0.441	0.367	0.417	0.664	0.874

4.1 Generation of regional financial behavior networks

The different mapping node rules are as follows:

(1) Considering that the capital situation in this article has significant difficulties in processing various data, each enterprise's capital situation will be based on the structural characteristics of the network, thereby establishing strategic cooperative relationships with the strength and characteristics of the bank's nodes. The mapping rules of each node are connected to the bank, thereby complicating the regional financial grid.

(2) When financial institutions such as banks act as providers of financing funds, different regional entities will use the financing model of accounts receivable and the core enterprises of the supply chain as guarantees. When evaluating the operation status of core enterprises in the supply chain, the financing qualification in the supply chain mainly depends on the core enterprises in the supply chain. In different regional financial grids, the nodes of each financial complex network are divided into two categories. According to certain rules, the funds of enterprises and banks are connected and established.

(3) Regarding all cooperation agreements signed by banks and trade transactions between enterprises as a form of financing, in actual operation, if each bank has a strict inspection system, it is necessary to inspect the actual situation and funds of all enterprises to ensure the reliability of repayment. In different enterprise nodes, if all grid nodes need to be mapped to a regional complex network, All high-quality enterprises need to be filtered.

(4) If the supply chain nodes of a region change with the nodes of the financial complex grid, the network of regional financial complexity will enter the financial grid. If the following nodes are mapped to two nodes on a certain edge, different regional financial grids will be connected to the bank nodes in the network.

In the enterprise selection ratio for each type of node, the number of banks varies according to the structural characteristics of the network, as shown in Table 3.

Table 3

Simulation results of regional financial grid
y	10	10	10
β2	0.15	0.25	0.35
Average degree	14.5711	18.9547	20.6871
Average path length	3.4158	3.3574	3.3485
Clustering coefficient	0.0048	0.0024	0.0055

The financial grid of a region grows as the number of nodes increases. In a complex network model, if the grid of regional finance changes as the hierarchy changes, it can indicate that both bank nodes and class C nodes have a certain connection effect. As nodes in the financial network join, the average path between different nodes will also shorten as the length increases. If the average value of the network increases significantly, many networks will change with the average path point along with the average degree of the financial grid. If the financial grids of different regions break the supply chain nodes of the original region, then the generated financial region grid coefficient is no longer 0.

Changing the number of banks in a regional financial grid will have a certain impact on all network characteristics. If the set number of banks is different, the resulting regional financial grid will be simulated based on relevant characteristics, as shown in Table 4:

Table 4

Simulation Results of Regional Financial Network Changing Only y
y	5	10	15
β2	0.25	0.25	0.25
Average degree	18.1125	18.6235	18.7812
Average path length	3.4236	3.5212	3.7951
Clustering coefficient	0.0012	0.0014	0.0021

Under the influence of complex grid indicators, if the structure of the financial grid changes due to changing the number of banks, it is necessary to compare the average number of banks before the study with the length of the simulated financial grid. If the average path length in the grid is relatively small, a major reason is that the number of enterprises is greater than the number of banks. If the impact of network structure on the number of banks is relatively small, then as the number of banks increases, it can be found that increasing the aggregation coefficient of the network will break the hierarchical nature of the original supply chain network. If the relationship between nodes becomes more complex, the specific coefficient will also increase.

4.2 Regional financial behavior simulation

According to the constraint conditions of the parameters, the replication dynamic stable points and parameters of enterprises and banks are simulated. According to the comparison of the software simulation and the change results of the main parameters in the equilibrium state, it can be found that the main parameters in the equilibrium state will change according to different values. If the size of each parameter and the main body of each zone represent a real data, then the parameter values of each zone will become the main relationship that affects the sensitivity of the factors. If the initial evolution time is set to 0, the simulation effect of the start and end points of time is shown in Fig. 5:

As can be seen from Fig. 5, the innovation of enterprises increases with the increase in the proportion of banks' selective lending strategies. If the probability of strict government regulation is increasing, it indicates that the innovation income of the original invention will change with the second innovation income. If the total revenue of a bank under strict government supervision is greater than the total revenue without supervision, the system will evolve into a good state. With the selective lending strategy of banks, enterprises will enter an ideal management mode after the original invention is re established, and it will not affect the speed of stable state and final evolution results of various regions.

With the increase of subsidies for enterprise innovation, major banks have a certain degree of benefit support for the innovation of enterprise technology and the coefficient of income. At the same time, the speed of positive equilibrium is also increasing for the research and development of enterprise innovative technology. In order to innovate the original inventions of major banks, the government has increased its management of the trend of bank equilibrium.

Due to the mismatch of financial resources nowadays, the cost of original invention and innovation for various enterprises has increased significantly, and many enterprises are developing towards secondary innovation. As time passes, the degree of mismatch becomes higher and its behavior tends to be more secondary innovation.

Each enterprise increases the property rights benefits of original inventions and innovations, reduces their risk costs, and increases the speed at which the benefits obtained by enterprises are directed towards state-owned and private enterprises. And under the influence of these factors, state-owned enterprises are not as obvious as private enterprises.

4.3 Regional financial behavior development model

By establishing a regional network platform on the internet, the economic operation has been online on a virtual network. Through this online approach, we have achieved regional economic growth. Through this online approach, combined with financial needs, we have conducted secure transactions between both parties on the network, resolving various time and space constraints associated with offline transactions. In terms of the advantages of online transactions, we can enable both parties or suppliers to conduct transactions and discussions on the same platform in various platforms or regions, so that the supplier or both parties to the transaction can learn about their characteristics, behavior, credit, and other aspects of information through online big data, thereby enabling both parties to complete online secure transactions. Platform technology can embody many advantages on the network, which facilitates the centralized management of financial regions, and thus improves the efficiency of financial region management.

By utilizing the characteristics of big data, sorting out massive data and information, and providing them based on user needs, it is convenient for users to meet various needs. With a large amount of data, a wide range of resources, and a complex range of knowledge on the network, the role of data can be of great help to users. Any data will be reflected in the economic system. Financial institutions will set up a financial risk early warning system on the network, which can predict possible risks in the financial system in advance through big data, so as to have sufficient time and energy to prevent and solve upcoming risks and obtain the priority and initiative to solve risks. On the other hand, it is very beneficial for financial institutions to conduct comprehensive and accurate evaluations of some enterprises, so as to facilitate the improvement of the level and quality of financial management. It can also provide the Party and the national government with economic decisions and programs that can reflect the economic operation of the relevant regional economy and finance, thereby providing a more objective basis.

The use of blockchain technology can facilitate financial business. This technology is a combination of many disciplines. It has many excellent features, and can also share information in real time. It can effectively solve information symmetry issues and improve the success rate and constraint rate of transactions. Blockchain technology can be very effective in transmitting information through regional financial information and maintaining its authenticity. Through the application of blockchain technology, participants in regional finance can easily understand the specific situation of each party. Participants can learn about each party's information in real time based on their permissions, and have a deep understanding of this information, ensuring that the information obtained by each party is consistent, accurate, authentic, and tamperproof. Because there is no hierarchical relationship in each region of the supply chain, which belongs to a parallel relationship, it is more conducive to cooperation between the two parties. Especially with the support of big data, all parties need to make appointments when participating in this matter. Therefore, big data will, based on the needs of all parties, enable the two parties to better perform their duties, and can directly pursue and adjudicate violations of contract performance, in order to maintain this order.

Using Internet of Things technology to assist regional financial services is a technology with many advantages. Through the Internet of Things technology, we can very accurately know the state and direction of things, and can also change the shape of things according to settings; Through the Internet of Things, we can also provide alarm prompts at any time, connect and cooperate with the police backstage, directly take measures against them, and cooperate with relevant judicial departments to take loss stop measures. Therefore, the Internet of Things technology not only solves various problems in regional financial products, but also makes a lot of contributions to the development of regional financial products.

Using artificial intelligence technology to assist regional financial business is a technology with many advantages. Artificial intelligence technology is of great significance for regional finance because the relationship between enterprises and the regional economy is very complex. To understand the situation of each enterprise, artificial intelligence cannot solve the problem by relying solely on human resources. However, using artificial intelligence technology can solve things that cannot be completed by human resources in a short time. Artificial intelligence can help various regional entities analyze and filter information in many ways, And draw corresponding conclusions in a very short time, improving the accuracy of many products in the operating area, as well as the control ability of many risks.

Coordination is the top priority of regional financial business development. It is necessary to have relevant professional institutions and organizations for reasonable management, as the economy of the relevant region is related to multiple entities, and the finance of the relevant region is related to multiple financial institutions. We can make full use of online platforms to communicate between industries, explore technologies and projects related to finance, enhance management within and outside the industry, so as to solve various problems in finance, and develop finance through these means. In our era of rapid development of network technology, facing the rapid development of network technology, it will appear helpless and unable to start. Recommendation algorithm is a very useful tool in the current era, which can provide the products and services required by users according to their needs. In the modern era of network enrichment, recommendation algorithms have also played a good role in it. The recommendation list is mainly based on the user's personal characteristics such as usage characteristics, hobbies, and characteristics, while traditional recommendations are mainly classified according to the user's characteristics and preferences. The use of references based on web based recommendation algorithms has also improved significantly, resulting in some very good results. After learning, machines can capture and analyze user information in the big data of the network, and simplify complex relationships. Therefore, they are deeply loved and favored by users. In summary, this article mainly studies the main uses of recommendation algorithms and promotes the development of regional financial behavior patterns.

Compliance with Ethical Standards

Conflict of interest

The authors declare that they have no conflict of interests

Ethical approval

This article does not contain any studies with human participants performed by any of the authors.

Data Availability

Data will be made available on request.

Agarwal S, Chomsisengphet S, Mahoney N, et al. Regulating consumer financial products: Evidence from credit cards[J]. The Quarterly Journal of Economics, 2015, 130(1): 111-164.
Keys B J, Mahoney N, Yang H. What determines consumer financial distress? Place-and person-based factors[J]. The Review of Financial Studies, 2023, 36(1): 42-69.
Zhu T, Xiao J J. Consumer financial education and risky financial asset holding in China[J]. International Journal of Consumer Studies, 2022, 46(1): 56-74.
Lin Z, Whinston A B, Fan S. Harnessing Internet finance with innovative cyber credit management[J]. Financial Innovation, 2015, 1: 1-24.
Guo L, Arnould E J, Gruen T W, et al. Socializing to co-produce: Pathways to consumers’ financial well-being[J]. Journal of Service Research, 2013, 16(4): 549-563.
Raymond D. Using artificial intelligence to combat information overload in research[J]. IEEE pulse, 2019, 10(1): 18-21.
Dietzmann C, Duan Y. Artificial intelligence for managerial information processing and decision-making in the era of information overload[C]//Proceedings of the 55th Hawaii International Conference on System Sciences. 2022.
Padmanabhan J, Johnson Premkumar M J. Machine learning in automatic speech recognition: A survey[J]. IETE Technical Review, 2015, 32(4): 240-251.
Musyoki D, Kadubo A S. The impact of credit risk management on the financial performance of banks in Kenya for the period[J]. International Journal of Business and Public Management, 2012, 2(2): 72-80.
Butaru F, Chen Q, Clark B, et al. Risk and risk management in the credit card industry[J]. Journal of Banking & Finance, 2016, 72: 218-239.
Chen Y, Kumara E K, Sivakumar V. Investigation of finance industry on risk awareness model and digital economic growth[J]. Annals of Operations Research, 2021: 1-22.
Park H, Kim J D. Transition towards green banking: role of financial regulators and financial institutions[J]. Asian Journal of Sustainability and Social Responsibility, 2020, 5(1): 1-25.
Phuong N T T, Huy D T N, Van Tuan P. The evaluation of impacts of a seven factor model on nvb stock price in commercial banking industry in vietnam-and roles of Discolosure of Accounting Policy In Risk Management[J]. International Journal of Entrepreneurship, 2020, 24: 1-13.
Wu S, Sun F, Zhang W, et al. Graph neural networks in recommender systems: a survey[J]. ACM Computing Surveys, 2022, 55(5): 1-37.
Gao C, Zheng Y, Li N, et al. Graph neural networks for recommender systems: Challenges, methods, and directions[J]. arXiv preprint arXiv:2109.12843, 2021.
Wu H, Zhang Z, Yue K, et al. Dual-regularized matrix factorization with deep neural networks for recommender systems[J]. Knowledge-Based Systems, 2018, 145: 46-58.
Kriegeskorte N, Golan T. Neural network models and deep learning[J]. Current Biology, 2019, 29(7): R231-R236.

Download PDF

Reviewers agreed at journal
11 Apr, 2023
Editor assigned by journal
30 Mar, 2023
First submitted to journal
29 Mar, 2023

You are reading this latest preprint version

Simulation of regional financial behavior patterns based on recommendation algorithm and machine learning

Status:

Version 1

Abstract

Figures

1 Introduction

2 Related Work

3 Principle Analysis Based On Recommendation Algorithm And Machine Learning

3.1 Theoretical basis of recommendation algorithm

3.2 Regional financial market prediction based on machine learning

4 Simulation Of Regional Financial Behavior Patterns

4.1 Generation of regional financial behavior networks

4.2 Regional financial behavior simulation

4.3 Regional financial behavior development model

5 Conclusion

Declarations

References

Status:

Version 1