Credit evaluation method and device based on artificial intelligence and big data technology

Abstract

The invention provides a credit assessment method and device based on artificial intelligence and a big data technology, and aims to improve the accuracy of credit risk assessment through the analysis of multi-dimensional behavior data. The method comprises the following steps: firstly, collecting multi-dimensional behavior data of a target customer, and extracting multi-dimensional behavior characteristics of the customer based on the data, including a behavior stability index, an abnormal behavior frequency, behavior trend data and multi-dimensional behavior correlation data; and generating a credit risk assessment result through a pre-trained neural network model in combination with the credit report data of the customer, the environmental risk assessment data and the like. And finally, through analysis of the credit risk assessment result, credit approval suggestions and personalized loan schemes for clients are generated. According to the invention, multi-dimensional data and macroeconomic factors are integrated in credit risk assessment, comprehensive assessment of customer credit is realized, and the accuracy of credit decision is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a credit assessment method and device based on artificial intelligence and big data technology. Background Art

[0002] In the existing technology, credit assessment usually relies on the historical credit record and financial data of the target customer for assessment, mainly through traditional credit reports, bank transaction records and other financial data to predict the customer's credit risk. These methods are often combined with simple scoring models or rule-based analysis systems to generate customer credit scores as an important basis for credit approval.

[0003] However, the main problem faced by existing technologies is that the data source for credit assessment is relatively single, mainly relying on the customer's financial history, and failing to fully utilize the customer's data in other behavioral dimensions.

[0004] Therefore, it is necessary to develop a new credit assessment method. Summary of the invention

[0005] The present application provides a credit assessment method and device based on artificial intelligence and big data technology to improve the accuracy of credit risk assessment.

[0006] This application provides a credit assessment method based on artificial intelligence and big data technology, including: Collect multi-dimensional behavior data of target customers, the multi-dimensional behavior data includes financial behavior data, social behavior data, consumption behavior data, geographic location behavior data and network behavior data; obtain multi-dimensional behavior characteristics of target customers based on the multi-dimensional behavior data, wherein the multi-dimensional behavior characteristics include the behavior stability index, abnormal behavior frequency, behavior trend data and multi-dimensional behavior correlation data of target customers; the behavior stability index is used to evaluate the consistency and volatility of target customers in each behavior dimension; the abnormal behavior frequency refers to the frequency of abnormal events in the behavior pattern of target customers that deviate from the normal state; the behavior trend data is used to reflect the change trend of the target customer's behavior over time and its development development speed; the multi-dimensional behavior correlation data is used to reflect the mutual correlation between different dimensional behavior data; based on the pre-trained neural network model, a credit risk assessment result is generated according to the target customer's credit report data, environmental risk assessment data, multi-dimensional behavior data and multi-dimensional behavior characteristics; wherein the credit risk assessment result includes a credit risk score, the identification result of key risk factors and a future credit risk forecast; the environmental risk assessment data includes economic indicator data, and the economic indicator data includes GDP growth rate, unemployment rate and industry risk data of the industry in which the target customer is engaged; based on the credit risk assessment result output by the neural network model, a credit approval recommendation and a personalized loan plan are generated.

[0007] In an optional embodiment, the neural network model includes a feature extraction network, an anomaly detection network and a multi-layer fusion network; The input of the feature extraction network is the multi-dimensional behavior data and multi-dimensional behavior features of the target customer; a convolutional autoencoder is used to extract features from the multi-dimensional behavior data of the target customer to obtain a nonlinear behavior feature vector; The input of the anomaly detection network is the nonlinear behavior feature vector provided by the feature extraction network; the anomaly detection network is implemented by a recursive neural network based on a multi-head self-attention mechanism to obtain abnormal behavior detection results and prediction probabilities of abnormal behaviors; The input of the multi-layer fusion network is the nonlinear behavior feature vector provided by the feature extraction network, the abnormal behavior detection results and the predicted probability of abnormal behavior provided by the anomaly detection network, the credit report data and the economic indicator data; the multi-layer fusion network is implemented using a capsule network to obtain the credit risk assessment results of the target customers.

[0008] In an optional implementation, the multi-layer convolutional autoencoder used in the feature extraction network includes: The first input layer is used to receive multi-dimensional behavior data and multi-dimensional behavior characteristics of target customers; A first convolutional layer, used to receive the multi-dimensional behavior data provided by the first input layer, wherein the first convolutional layer uses 64 3×3 convolution kernels to perform a convolution operation to extract local features, and outputs 64 feature maps; The first pooling layer is used to receive the output of the first convolutional layer, use a 2×2 pooling window to perform a maximum pooling operation, reduce the spatial dimension of the feature map, and output 64 pooled feature maps; The second convolution layer is used to receive the output of the first pooling layer. The second convolution layer uses 128 3×3 convolution kernels to extract deeper local features, capture temporal dependencies and nonlinear relationships, and outputs 128 feature maps. The second pooling layer is used to receive the output of the second convolutional layer. The second pooling layer uses a 2×2 pooling window to perform a maximum pooling operation to further reduce the spatial dimension of the features, and the output is 128 pooled feature maps; The encoding layer receives the output of the second pooling layer and compresses the feature vector into a 256-dimensional high-dimensional nonlinear behavior feature vector through a fully connected layer as the intermediate result of the feature extraction network and used in the decoding stage. A first deconvolution layer is used to receive the 256-dimensional feature vector output by the encoding layer. The first deconvolution layer uses 128 3×3 convolution kernels to perform a deconvolution operation, restore part of the spatial dimensions, and output 128 feature maps; The first upsampling layer is used to receive the output of the first deconvolution layer, perform a 2×2 upsampling operation, restore the spatial resolution of the data, and generate 128 enlarged feature maps; A second deconvolution layer is used to receive the output of the first upsampling layer. The second deconvolution layer uses 64 3×3 convolution kernels to further perform a deconvolution operation to restore a higher spatial resolution and output 64 feature maps. A second upsampling layer, used to receive the output of the second deconvolution layer, the second upsampling layer performs a 2×2 upsampling operation to restore the spatial resolution to a similar spatial resolution as the first convolution layer, and outputs 64 feature maps; The first output layer is used to further process the output of the second upsampling layer through a convolutional layer to generate reconstructed data similar to the first input layer; and to generate a 256-dimensional high-dimensional nonlinear behavioral feature vector to represent the abstract behavioral characteristics of the target customers.

[0009] In an optional implementation, the anomaly detection network uses a recursive neural network based on a multi-head self-attention mechanism, including: The second input layer is used to receive the nonlinear behavior feature vector output from the feature extraction network; The spatiotemporal graph convolution layer is used to receive the nonlinear behavior feature vector provided by the second input layer. The spatiotemporal graph convolution layer constructs a spatiotemporal graph based on the graph convolution network and the time dimension, and is specifically used for: By arranging the target customers' multi-dimensional behavior data and multi-dimensional behavior characteristics in chronological order, a spatiotemporal graph is formed, in which nodes represent the behavior states at different time points, and edges represent the behavior relationships between different time points; The graph convolutional network is used to convolve the nodes in the spatiotemporal graph to capture the dependency of user behavior in the time dimension and the spatial characteristics of behavior data, and obtain a 128-dimensional feature vector after spatiotemporal convolution. The multi-head self-attention layer is used to receive the 128-dimensional feature vector output from the spatiotemporal graph convolution layer, and use the multi-head self-attention mechanism to perform weighted processing on the key behavior patterns to generate a 1024-dimensional feature vector; The recursive neural network layer receives the 1024-dimensional feature vector from the multi-head self-attention layer, combines it with the long short-term memory network to process the time series characteristics of the behavior data, and outputs a 512-dimensional time series feature vector to represent the time characteristics of the abnormal behavior; The second output layer is used to receive the 512-dimensional time series feature vector provided by the recurrent neural network layer to generate the abnormal behavior prediction probability and abnormal behavior detection results of the target customers.

[0010] In an optional implementation, the multi-layer fusion network is implemented using a dynamic routing capsule network, and the dynamic routing capsule network includes: The third input layer is used to receive the nonlinear behavior feature vector provided by the feature extraction network, the abnormal behavior detection results and abnormal behavior prediction probability provided by the anomaly detection network, the credit report data, and the economic indicator data; the input data of the third input layer is integrated into a 919-dimensional feature vector as the input of the subsequent capsule layer; The first capsule layer is used to receive the 919-dimensional feature vector of the third input layer, and process different types of input features through multiple capsule units, wherein each capsule unit contains 8 neurons, which are responsible for processing data of different dimensions; the output of each capsule unit is an 8-dimensional vector, which represents the feature information of the input feature in different dimensions; and after passing through the first capsule layer, the output is a 64-dimensional feature vector; the dynamic routing mechanism of the first capsule layer includes: dynamically calculating the information transmission weight between capsule units according to the output of the capsule unit; the vector length of the capsule unit output represents the feature strength, and dynamically adjusting the information transmission path between different capsule units to strengthen the capsule output related to the key features; The second capsule layer is used to receive the 64-dimensional vector from the first capsule layer, and further fuse and process feature information from different sources; the second capsule layer includes multiple first capsule units, wherein each first capsule unit includes 16 neurons, and is used to process multi-dimensional features; the second capsule layer divides the 64-dimensional feature vector and inputs it into each capsule unit through a dynamic routing mechanism; the output of the second capsule layer is a 128-dimensional feature vector, which reflects the deep correlation between the target customer behavior data, economic data and anomaly detection results; A feature fusion layer is used to receive the 128-dimensional feature vector output by the second capsule layer and perform feature fusion through a fully connected layer; the output of the feature fusion layer is a 64-dimensional fused feature vector; The third output layer is used to receive the 64-dimensional fused feature vector from the feature fusion layer and generate a comprehensive credit risk assessment result of the target customer. The third output layer is specifically used to: Through the third fully connected layer, the 64-dimensional feature vector is mapped into a credit risk score, which represents the credit risk level of the target customer; The fourth fully connected layer is used to identify and output key factors that affect the customer's credit risk, and the output is a set of classification labels or specific values; wherein the key factors include abnormal behavior, credit history, and changes in the economic environment; Through the time series analysis layer, the 64-dimensional feature vector is combined with historical behavior data for prediction to generate the credit risk change trend for a specified period in the future, and the output is a future risk score that represents the customer's potential credit risk in the future.

[0011] In an optional implementation, the behavior stability index is calculated by a time-weighted average method, specifically including: Time slice the multi-dimensional behavior data and divide the behavior data into different time windows; Calculate the fluctuation range of the behavioral characteristics of each dimension within each time window; The behavioral fluctuations of each time window are weighted averaged, where more recent time windows are given greater weights, to assess the stability of the target customer’s current behavior and generate a behavioral stability index.

[0012] In an optional implementation, the abnormal behavior frequency detection is implemented by threshold setting and adaptive learning mechanism based on historical behavior patterns, including: Establish a behavioral baseline pattern by analyzing the target customers’ behavioral data over the past 6 months to 1 year; Setting a threshold for deviation from historical behavior patterns, and behavioral events exceeding this threshold are marked as abnormal; By applying an adaptive learning algorithm, the threshold is updated in real time based on the new behavior data of target customers, making the model more accurate in detecting abnormal behavior.

[0013] In an optional implementation, the calculation of the multi-dimensional behavior correlation data adopts a method based on the Pearson correlation coefficient, specifically including: Normalize the behavioral characteristic data of target customers in different dimensions; Calculate the correlation between behavioral data of various dimensions, including financial behavior and consumer behavior, social behavior and geographic location behavior, etc.; The Pearson correlation coefficient is used to evaluate the linear correlation between behavioral data of different dimensions, generate a correlation matrix, and use it as input for subsequent credit risk assessment models.

[0014] In an optional implementation, the acquisition and update of the environmental risk assessment data is achieved through a dynamic macroeconomic monitoring system based on big data, specifically including: Obtain the latest GDP growth rate, unemployment rate and industry risk data from government statistics departments, industry reports and third-party economic data providers; Regularly update the economic indicators of the industries in which target customers are engaged, and dynamically adjust the credit risk of target customers in light of macroeconomic environments such as economic recession and rising industry risks.

[0015] The present application also provides a credit assessment device based on artificial intelligence and big data technology, including: A collection unit, used to collect multi-dimensional behavior data of target customers, wherein the multi-dimensional behavior data includes financial behavior data, social behavior data, consumption behavior data, geographic location behavior data, and network behavior data; An obtaining unit is used to obtain the multi-dimensional behavior characteristics of the target customer according to the multi-dimensional behavior data, wherein the multi-dimensional behavior characteristics include the behavior stability index, abnormal behavior frequency, behavior trend data and multi-dimensional behavior correlation data of the target customer; the behavior stability index is used to evaluate the consistency and volatility of the target customer in each behavior dimension; the abnormal behavior frequency refers to the frequency of abnormal events that deviate from the normal state in the behavior pattern of the target customer; the behavior trend data is used to reflect the change trend of the target customer's behavior over time and its development speed; the multi-dimensional behavior correlation data is used to reflect the mutual correlation between the behavior data of different dimensions; A generating unit, for generating a credit risk assessment result based on a pre-trained neural network model and according to the target customer's credit report data, environmental risk assessment data, multi-dimensional behavior data and multi-dimensional behavior characteristics; wherein the credit risk assessment result includes a credit risk score, identification results of key risk factors and future credit risk forecasts; the environmental risk assessment data includes economic indicator data, and the economic indicator data includes GDP growth rate, unemployment rate and industry risk data of the industry in which the target customer is engaged; The approval unit is used to generate credit approval suggestions and personalized loan plans based on the credit risk assessment results output by the neural network model.

[0016] The technical solution proposed in this application has the following beneficial technical effects: (1) By collecting multi-dimensional behavioral data of target customers (including financial, social, consumption, geographic location and network behavior data), the present invention can comprehensively evaluate the credit status of customers from multiple aspects, providing more accurate credit risk assessment results compared to relying solely on traditional financial data. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flowchart of a credit assessment method based on artificial intelligence and big data technology provided in the first embodiment of the present application. DETAILED DESCRIPTION

[0018] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.

[0019] The first embodiment of the present application provides a credit assessment method based on artificial intelligence and big data technology. Figure 1 , which is a schematic diagram of the first embodiment of the present application. Figure 1 A credit assessment method based on artificial intelligence and big data technology is described in detail in the first embodiment of the present application.

[0020] Step S101: Collect multi-dimensional behavior data of target customers, wherein the multi-dimensional behavior data includes financial behavior data, social behavior data, consumption behavior data, geographic location behavior data, and network behavior data.

[0021] It should be pointed out that the multi-dimensional behavior data here does not necessarily include behavior data of all dimensions, such as financial behavior data, social behavior data, consumption behavior data, geographic location behavior data, and network behavior data, but may also be one or more of these dimensional data.

[0022] First of all, financial behavior data can be collected through cooperation with financial institutions, third-party payment platforms or data providers authorized by the customer to obtain information such as the target customer’s bank account information, credit card transaction records, loan repayment history, investment behavior, etc.

[0023] Secondly, the collection of social behavior data requires explicit authorization from users and is conducted through legitimate third-party data service providers. This type of data mainly comes from social networking platforms, analyzing users’ interactive behaviors, friend relationships, changes in social circles, and activity on the platform.

[0024] Consumption behavior data is mainly collected through customers' shopping records, online consumption, etc. This can be obtained through the data interfaces of e-commerce platforms and payment platforms. The data types include shopping categories, consumption frequency, consumption amount, payment methods and other information. These data can reflect customers' consumption habits, spending stability and financial health. For large e-commerce platforms, customers' consumption data can be directly obtained through the platform API and combined with the payment platform data authorized by the user to provide a more detailed consumption portrait.

[0025] Geographic location behavior data is collected through the GPS information of mobile devices, payment location records, and other location-related services. This data mainly reflects the target customers' daily travel patterns, changes in work and residence locations, and travel frequency. Data sources include navigation applications used by customers, location information records of payment platforms, and other authorized data provided by mobile operators.

[0026] The collection of online behavior data is achieved through cooperation with third-party data providers authorized by users to obtain target customers' online search history, types of browsed websites, visit frequency, and online time distribution.

[0027] Step S102: According to the multi-dimensional behavior data, the multi-dimensional behavior characteristics of the target customer are obtained, wherein the multi-dimensional behavior characteristics include the target customer's behavior stability index, abnormal behavior frequency, behavior trend data and multi-dimensional behavior correlation data; the behavior stability index is used to evaluate the consistency and volatility of the target customer in each behavior dimension; the abnormal behavior frequency refers to the frequency of abnormal events that deviate from the normal state in the target customer's behavior pattern; the behavior trend data is used to reflect the change trend of the target customer's behavior over time and its development speed; the multi-dimensional behavior correlation data is used to reflect the correlation between behavior data of different dimensions.

[0028] Step S102 involves extracting multi-dimensional behavioral features of target customers based on multi-dimensional behavioral data. These behavioral features include behavioral stability index, abnormal behavior frequency, behavioral trend data, and multi-dimensional behavioral correlation data. The generation process of each feature requires detailed calculation and analysis to ensure its accuracy and effectiveness in assessing the credit risk of target customers.

[0029] First, the calculation of the behavioral stability index is achieved by analyzing the changes in the behavioral data of target customers in different time periods. Specifically, a time series analysis is performed on each type of behavioral data (such as financial behavior, social behavior, consumer behavior, etc.), which is divided into several fixed time windows (such as daily, weekly or monthly). In each time window, the volatility of each behavioral data is calculated, and statistical methods (such as variance, standard deviation) are used to measure the amplitude of these fluctuations. Next, based on the volatility data in these time windows, the time-weighted average is calculated, and a higher weight is given to data closer to the current time to reflect the stability of the customer's recent behavior. In this way, a behavioral stability index that reflects the consistency and volatility of customer behavior is obtained, which is used to evaluate the credit stability of customers.

[0030] Secondly, the detection of abnormal behavior frequency is achieved by establishing a baseline for the behavior pattern and continuously monitoring it. First, collect the historical behavior data of the target customers over a period of time (for example, 6 months) to establish a baseline behavior pattern. These baseline data are used to set thresholds for deviations from normal behavior. When the customer's new behavior data exceeds the set threshold compared to the historical baseline, these behaviors will be marked as abnormal. The system generates abnormal behavior frequency data by counting the frequency of these abnormal behaviors. To improve accuracy, machine learning algorithms can be used to dynamically adjust the threshold to ensure that the system can adapt to changes in customer behavior over time and identify abnormal behavior patterns.

[0031] For the generation of behavioral trend data, a time series analysis method is used to record the changes in the target customer's behavior in different time periods. By performing differential processing on the behavioral data of each time period, the rate and direction of change of the behavior are calculated. This step can not only identify the overall trend of customer behavior, but also capture the speed and magnitude of behavioral changes. For example, consumption habits in financial behavior may increase or decrease sharply in a short period of time. The system generates corresponding behavioral trend data by analyzing these change rates and is used for further assessment of credit risk.

[0032] Finally, the calculation of multi-dimensional behavioral correlation data is achieved through statistical methods and correlation analysis tools. The specific steps are to normalize the customer's behavioral data in different dimensions and then calculate the correlation between these data. For example, the relationship between financial behavior and consumer behavior, geographic location behavior and social behavior, etc. can be analyzed through the Pearson correlation coefficient or other common correlation calculation methods. These correlation data are used to evaluate the interactive impact between various behaviors, thereby providing more dimensional information support for subsequent credit risk assessment.

[0033] Through the above steps, the system can effectively extract the multi-dimensional behavioral characteristics of customers, which will be used in subsequent neural network models to generate more accurate credit risk assessment results.

[0034] Furthermore, the behavior stability index is calculated by the following formula: ; According to the formula, the Behavior Stability Index (BSI) is a complex calculation that combines multi-dimensional data to quantify the behavioral stability of target customers. The following is a detailed description of each parameter in Formula 1.

[0035] 1. BSI stands for Behavioral Stability Index: BSI represents the comprehensive behavioral stability of customers in all behavioral dimensions. It evaluates the stability of customer behavior by combining behavioral data, volatility, time dependency and other factors in each dimension. The higher the BSI value, the more stable the customer's behavior; conversely, a lower value indicates that the customer's behavior fluctuates greatly or has abnormal changes.

[0036] 2. Indicates Dimensional behavioral data: Representing target customers Behavioral data on dimensions, such as financial behavior, consumer behavior, social behavior, etc. Behavioral data on each dimension is obtained by monitoring the customer's daily activities or transaction behaviors. Specifically, financial behavior may include income and expenditure, and consumer behavior may include shopping frequency and amount. The data of each dimension is taken as an independent input.

[0037] 3. and Represents the historical mean and historical standard deviation: It is The historical mean of the behavioral data on a dimension is obtained by long-term observation of the customer's behavioral data on this dimension, and represents the long-term average behavior pattern of the customer on this dimension.

[0038] It is The standard deviation of the behavioral data on a dimension indicates the volatility or dispersion of the behavioral data, reflecting the magnitude of the customer's behavioral changes on this dimension. A higher standard deviation means that the customer's behavior on this dimension is unstable and fluctuates greatly, while a lower standard deviation means that the behavior is relatively stable.

[0039] 4. and Indicates adjustment parameters: and These are parameters used to adjust the impact of the deviation and standard deviation of behavioral data. These parameters are dynamically adjusted according to specific application requirements to increase or decrease the weight of certain behavioral features in the overall stability index.

[0040] What is adjusted is the behavioral data Deviation from historical mean The weight of , which reflects the degree of difference between the customer's current behavior and past behavior. The larger it is, the greater the impact on the deviation.

[0041] is the standard deviation The weight adjustment controls the impact of historical fluctuations on the stability index. This means that the standard deviation has a greater impact, that is, behavioral volatility has a more significant impact on the stability of this dimension.

[0042] 5. Represents time weight: It is The time weight of dimension behavior data is used to reflect the time dependency of the customer's behavior in this dimension. This weight is usually based on the customer's behavior history and is obtained by analyzing the trend of behavior changes over time. Indicates the behavior pattern of customers over a specific period of time, for example, may indicate the long-term stability of the customer's behavior on this dimension, while a smaller This may mean that the behavior of this dimension has shorter-term volatility.

[0043] 6. Denotes the fluctuation correction factor: It is The fluctuation correction factor of the dimension describes the fluctuation frequency or intensity of the behavior dimension. By performing frequency analysis on the historical behavior data of this dimension, we can obtain .

[0044] For example, if the customer's behavior changes frequently on this dimension, the fluctuation correction factor will be larger, reflecting the instability of behavior. It will be smaller, indicating that customer behavior on this dimension is relatively stable.

[0045] 7. Indicates the time interval: It is The time interval between the latest behavior data and the previous behavior data on the dimension. It is used to measure the timeliness of customer behavior changes. If the behavior interval is short (i.e., customer behavior changes frequently), then smaller; conversely, longer intervals indicate sparser behavioral changes.

[0046] 8. and Indicates control parameters: and It is a parameter that controls time dependence and fluctuation, and is used to dynamically adjust the impact of time intervals on the stability of customer behavior.

[0047] Controlling the time-dependent sensitivity, higher The time interval of the behavioral data has a greater impact on stability, while Used to offset the time interval and adjust the nonlinear effect of time changes on behavioral stability.

[0048] 9. Represents weight: It is The weight of the dimension behavior data, which indicates the contribution of each dimension to the overall behavior stability index. The weight is usually set based on business needs or the importance of the dimension in the customer's overall credit risk. For example, financial behavior may be more important than social behavior for credit risk assessment, so the weight of financial behavior is Will be relatively high.

[0049] 10. Indicates the total number of dimensions: It is the total number of dimensions of customer behavior data, indicating how many different types of behavior data are analyzed by the system. Usually these dimensions include financial behavior, consumer behavior, social behavior, geographic location behavior, and network behavior.

[0050] Through this formula, the system can take into account the volatility, time dependence, weight and other factors of each behavioral dimension and dynamically calculate the customer's behavioral stability index (BSI). This approach ensures the comprehensive quantification of customer behavior, especially in complex and multi-dimensional scenarios, and can capture subtle changes in customer behavior and ensure the accuracy of the evaluation.

[0051] Furthermore, the calculation of the behavior stability index is achieved by a time-weighted average method, specifically including: Time slice the multi-dimensional behavior data and divide the behavior data into different time windows; Calculate the fluctuation range of the behavioral characteristics of each dimension within each time window; The behavioral fluctuations of each time window are weighted averaged, where more recent time windows are given greater weights, to assess the stability of the target customer’s current behavior and generate a behavioral stability index.

[0052] First, the processing of behavioral data is to divide multi-dimensional behavioral data (including financial behavior, social behavior, consumer behavior, etc.) into fixed time intervals through time slicing technology. Specifically, these time windows can be set to daily, weekly or monthly according to actual needs. Each time window contains all the behavioral data of the target customer during that time period, thus forming behavioral information with a time series. This time slicing process ensures that subsequent analysis can identify the time dependency of behavioral data and lays the foundation for the calculation of the behavioral stability index.

[0053] Next, the system analyzes the behavioral data within each time window and calculates the fluctuation range of each behavioral feature. The fluctuation range reflects the magnitude of changes in the customer's behavioral data within a specific time window. Specifically, the calculation of the fluctuation range can be achieved through common statistical methods such as standard deviation or variance to measure the degree of fluctuation of a behavior within the time window. For different behavioral dimensions (such as fluctuations in the amount of consumption behavior, changes in the frequency of financial transactions, etc.), the system will calculate their fluctuation ranges separately in order to comprehensively measure changes in customer behavior.

[0054] After calculating the behavioral fluctuations in each time window, the system will perform a weighted average of these fluctuation data. The core of weighted averaging is to assign different weights to data in different time windows. Since the customer's recent behavior is more valuable for the current credit risk assessment, the more recent time windows are usually given a larger weight. Although relatively distant time windows are also involved in the calculation, their weights are relatively low. The distribution of weights can be automatically adjusted according to business rules or through algorithms, making the calculation results more flexible and able to adapt to the actual fluctuations in customer behavior. Such a weight design helps to improve the accuracy of the behavioral stability index, especially when the customer's recent behavior fluctuates greatly, it can reflect this change more promptly.

[0055] Finally, the system will generate a behavior stability index based on the results of these weighted calculations. This index comprehensively considers the fluctuations in customer behavior in each time window and time factors, and quantifies the overall stability of customer behavior. As an important credit assessment indicator, the behavior stability index can help credit institutions better judge the credit risk of customers, especially in assessing the consistency and short-term volatility of customer behavior.

[0056] Furthermore, the detection of abnormal behavior frequency is achieved through threshold setting and adaptive learning mechanism based on historical behavior patterns, including: Establish a behavioral baseline pattern by analyzing the target customers’ behavioral data over the past 6 months to 1 year; Setting a threshold for deviation from historical behavior patterns, and behavioral events exceeding this threshold are marked as abnormal; By applying an adaptive learning algorithm, the threshold is updated in real time based on the new behavior data of target customers, making the model more accurate in detecting abnormal behavior.

[0057] First, the system needs to collect and analyze the target customers' behavioral data, which usually involves multi-dimensional behavioral data, such as financial transactions, consumer behavior, social interactions, and geographic location changes. By conducting in-depth analysis of these behavioral data over the past 6 months to 1 year, the system can identify the customer's regular behavior patterns. This baseline pattern not only reflects the characteristics of the customer's daily behavior, but also includes the fluctuation range and frequency of these behaviors. For example, a customer's financial transactions may have a fixed periodicity and range, consumption habits remain stable within a specific category and amount range, and the frequency and intensity of social activities also show a certain regularity. By analyzing these data, the system can form a comprehensive, quantified behavioral baseline.

[0058] After establishing the baseline pattern, the system sets a threshold for determining abnormal behavior. The threshold represents the permissible range of deviation of the customer's current behavior from its historical pattern. If the customer's behavior exceeds this preset threshold, the system will mark the behavior as abnormal. For example, if the customer's spending amount suddenly increases significantly, exceeding the preset range of his normal spending pattern, the system will mark this event as abnormal behavior. The setting of thresholds is usually achieved through statistical analysis, and the degree of deviation can be defined based on the variance or standard deviation of the historical behavior data. A higher variance may indicate that the customer's behavior fluctuates greatly, so the threshold will be correspondingly higher. For customers with more stable behavior, the threshold will be relatively low to capture subtle anomalies.

[0059] During the abnormal behavior detection process, an adaptive learning algorithm is applied to dynamically adjust this threshold. Over time, customer behavior may change, especially when affected by external circumstances or changes in personal life. To ensure that abnormal behavior detection is always accurate, the system updates the threshold in real time through an adaptive learning mechanism. Specifically, as new behavior data is continuously input, the system automatically adjusts the deviation threshold based on the comparison of the new data with the historical baseline. If the customer's behavior gradually changes, such as a shift in consumption habits, the system can gradually increase or decrease the threshold by learning from this trend, so that the detection of abnormal behavior remains accurate. This adaptive learning mechanism ensures that the model can flexibly respond to dynamic changes in customer behavior and avoid overly frequent or insufficient abnormal behavior alerts.

[0060] The system ensures accurate detection of abnormal behavior by establishing historical behavior patterns, dynamically setting thresholds, and applying adaptive learning mechanisms. This method can not only quickly identify abnormal events in customer behavior, but also continuously optimize detection standards as customer behavior changes, thereby providing reliable data support for credit risk assessment.

[0061] Furthermore, the calculation of the multi-dimensional behavior correlation data adopts a method based on the Pearson correlation coefficient, which specifically includes: Normalize the behavioral characteristic data of target customers in different dimensions; Calculate the correlation between behavioral data of various dimensions, including financial behavior and consumer behavior, social behavior and geographic location behavior, etc.; The Pearson correlation coefficient is used to evaluate the linear correlation between behavioral data of different dimensions, generate a correlation matrix, and use it as input for subsequent credit risk assessment models.

[0062] First, the system normalizes the multi-dimensional behavioral data of target customers. This step is to ensure that data of different dimensions are comparable, because different types of behavioral data usually have orders of magnitude differences. For example, financial behavioral data may involve large transaction amounts, while social behavioral data may be related to interaction frequency. Before these data are normalized, they may cause deviations when calculating correlations. Therefore, normalization can scale the data of each behavioral dimension to the same range, making them consistent when calculating correlations, ensuring that each behavioral feature is reasonably considered.

[0063] Next, the system calculates the correlation between the behavioral data of each dimension. The different dimensions here usually include financial behavior, consumer behavior, social behavior, geographic location behavior and network behavior. By comparing and analyzing these behavioral data, the system can identify potential connections between them. For example, the correlation between financial behavior and consumer behavior may reflect the relationship between the customer's financial status and consumption habits; while the correlation between social behavior and geographic location behavior may reveal the relationship between the customer's social activities and changes in their geographic location. The relationship between these dimensions is important for assessing the customer's overall credit risk because they can provide additional contextual information to help identify hidden risk patterns.

[0064] In the correlation calculation, the Pearson correlation coefficient is used to evaluate the linear relationship between two behavioral dimensions. The Pearson correlation coefficient is a common statistical method used to measure the degree of linear correlation between two variables, and its value range is between -1 and 1. A coefficient of 1 indicates that the two variables are completely positively correlated, -1 indicates a complete negative correlation, and 0 indicates no correlation. In this method, the system will calculate the Pearson correlation coefficient between behavioral data of different dimensions pair by pair. For example, the system will calculate the correlation coefficient between financial behavior and consumer behavior, the correlation coefficient between social behavior and geographic location behavior, and so on. Each calculation result reflects the strength of the linear relationship between behaviors of different dimensions.

[0065] Finally, the system generates a correlation matrix based on the calculated Pearson correlation coefficient. Each element in the matrix represents the correlation between two types of behavioral data, and the relationship between all dimensional behaviors is intuitively presented in the form of a matrix. This correlation matrix not only provides important reference data for credit assessment, but also serves as an input for subsequent credit risk assessment models, helping the system to comprehensively consider the multi-dimensional behavioral characteristics of customers when conducting credit assessments.

[0066] This method combines data normalization, correlation calculation and the application of the Pearson correlation coefficient to ensure accurate analysis of the correlation between customers in various behavioral dimensions, providing a solid data foundation for credit risk assessment.

[0067] Furthermore, the acquisition and update of the environmental risk assessment data is achieved through a dynamic macroeconomic monitoring system based on big data, specifically including: Obtain the latest GDP growth rate, unemployment rate and industry risk data from government statistics departments, industry reports and third-party economic data providers; Regularly update the economic indicators of the industries in which target customers are engaged, and dynamically adjust the credit risk of target customers in light of macroeconomic environments such as economic recession and rising industry risks.

[0068] First, the system obtains the latest macroeconomic data from a variety of sources. These data sources include government statistical departments, industry reports, and third-party economic data providers. Specifically, government statistical departments usually publish authoritative economic indicators, such as GDP growth and unemployment rates for countries or regions. These data directly reflect the overall economic health. GDP growth can be used as a sign of economic growth or contraction, while unemployment rate reflects the supply and demand of the labor market. Industry reports provide detailed analysis of a specific industry, including current market demand, technological innovation, and competition. This type of information is particularly useful for assessing the risk level of the industry in which the target customer is engaged. Third-party economic data providers also provide a large amount of real-time updated economic data, covering global and regional economic trends, policy changes and other key information. These data sources are integrated into the system through API interfaces or regular downloads to ensure the timeliness and accuracy of the data.

[0069] After data collection is completed, the system processes these economic indicators and analyzes the specific industries in which the target customers are engaged. At this point, the system does not just stay at the general macroeconomic level, but goes deep into the specific economic environment of the customer's industry. For example, an industry may face the risk of economic recession, uncertainty caused by technological changes, or market shrinkage due to supply chain problems. These industry-specific economic factors will directly affect the customer's credit risk. Therefore, the system regularly updates relevant economic data based on the industry in which each customer is engaged, and analyzes the possible impact of economic fluctuations within and outside the industry on the customer's future credit performance.

[0070] After obtaining and analyzing these macroeconomic and industry data, the system will dynamically adjust the customer's credit risk assessment results. Specifically, the system will automatically adjust the customer's credit risk score based on changes in economic indicators, such as slowing GDP growth, rising unemployment, or increased industry risks. For example, if the target customer's industry is severely impacted by the economic recession, the system will increase its credit risk score accordingly, indicating the potential possibility of default. On the contrary, in the case of economic recovery or optimistic industry prospects, the system will appropriately lower the customer's credit risk score, reflecting its sound credit status in a good economic environment.

[0071] Step S103: Based on the pre-trained neural network model, a credit risk assessment result is generated according to the target customer's credit report data, environmental risk assessment data, multi-dimensional behavior data and multi-dimensional behavior characteristics; wherein the credit risk assessment result includes a credit risk score, identification results of key risk factors and future credit risk forecasts; the environmental risk assessment data includes economic indicator data, and the economic indicator data includes GDP growth rate, unemployment rate and industry risk data of the industry in which the target customer is engaged.

[0072] Step S103 involves generating the credit risk assessment results of the target customers based on the pre-trained neural network model. In this process, the system needs to take the target customers' multi-dimensional behavior data, behavior characteristics, credit report data and environmental risk assessment data as input, and obtain the final credit risk assessment results after calculation by the neural network model. The specific implementation of this step requires multi-level data input and complex deep learning model to coordinate processing to ensure that the output results are sufficiently accurate and comprehensive.

[0073] Furthermore, the neural network model includes a feature extraction network, an anomaly detection network and a multi-layer fusion network; First, the feature extraction network is responsible for processing the multi-dimensional behavioral data and multi-dimensional behavioral features of the target customers. Multi-dimensional behavioral data usually includes various types of information such as finance, consumption, social, geographic location, and network behavior. In order to extract key features from these complex data, the feature extraction network adopts the structure of convolutional autoencoder. Convolutional autoencoder consists of two parts: encoder and decoder. The encoder part gradually compresses the input raw behavioral data through multi-layer convolution operations, extracts representative local features, and represents this information as a low-dimensional feature vector. This process helps to remove noise from the data and retain important behavioral patterns. Next, the decoder reconstructs the compressed feature vector through a reverse convolution operation to evaluate whether the encoding accurately captures the key information in the input data. Finally, the output of the feature extraction network is a nonlinear behavioral feature vector, which is a highly abstract feature representation that can reflect the complexity and internal connections of customer behavior.

[0074] Next, the task of the anomaly detection network is to identify abnormal behaviors based on the nonlinear behavioral feature vectors provided by the feature extraction network. The core of this network uses a recursive neural network (RNN with Multi-head Self-Attention) based on a multi-head self-attention mechanism. In this network structure, the recursive neural network is responsible for processing time series data and identifying anomalies in behavior patterns by capturing the temporal changes in the target customer's behavior. The output of the network includes two key results: abnormal behavior detection results, which indicate behavioral events discovered by the system that are inconsistent with the customer's normal behavior pattern; and the predicted probability of abnormal behavior, which indicates the possibility of abnormal behavior by the customer again in the future. These outputs provide an important reference for subsequent credit risk assessment.

[0075] Finally, the multi-layer fusion network integrates inputs from multiple data sources, including the nonlinear behavior feature vectors of the feature extraction network, the abnormal behavior detection results and prediction probabilities of the anomaly detection network, the credit report data, and the economic indicator data. The multi-layer fusion network adopts the Capsule Network structure. The capsule network can better maintain the interdependence and spatial relationship between different input features by processing the multidimensional features and hierarchical relationships of the data. Each capsule unit processes a different set of features and passes important information to the next layer through a dynamic routing mechanism to ensure that key information is not lost when processing complex multi-source data. Finally, the capsule network generates the credit risk assessment results of the target customer, which not only includes the customer's current credit status score, but also provides possible future credit risk change trends by combining economic environment data and abnormal behavior predictions.

[0076] Furthermore, the multi-layer convolutional autoencoder used in the feature extraction network specifically includes: The first input layer receives the target customer’s multi-dimensional behavior data and multi-dimensional behavior characteristics. The multi-dimensional behavior data may include the customer’s financial behavior, consumption behavior, social behavior, etc. Through this input layer, all data is uniformly input into the network as the basic data for subsequent convolution operations.

[0077] In the first convolutional layer, the system uses 64 3×3 convolution kernels to perform convolution operations on the input multi-dimensional behavioral data. The function of convolution is to extract local features in the behavioral data, such as local temporal patterns or correlations between behaviors. The convolution operation generates 64 feature maps, which are preliminary abstract representations of the original data and extract the basic patterns in the data.

[0078] Next, the first pooling layer receives the output of the first convolutional layer and performs a maximum pooling operation using a 2×2 pooling window. The role of pooling is to reduce the spatial dimension of the feature map, that is, to compress the data, making subsequent processing more efficient while retaining the most important features. The pooling operation outputs 64 pooled feature maps, which retain the main information of local features while reducing data redundancy.

[0079] Subsequently, the second convolutional layer receives the output of the first pooling layer and uses 128 3×3 convolution kernels to perform a deeper convolution operation. Unlike the first convolutional layer, the purpose of the second convolutional layer is to capture more complex temporal dependencies and nonlinear behavior characteristics. The convolution of this layer can identify deeper feature patterns, thereby generating 128 new feature maps.

[0080] The second pooling layer receives the output of the second convolutional layer and uses a 2×2 pooling window to perform the maximum pooling operation again to further reduce the spatial dimension of the feature and output 128 pooled feature maps. This step ensures that the feature map is further compressed while retaining key information, laying the foundation for subsequent encoding processing.

[0081] The encoding layer receives the output of the second pooling layer and compresses the feature vector to 256 dimensions through a fully connected layer. This step further compresses the high-dimensional behavioral features into a high-dimensional nonlinear vector. This 256-dimensional vector contains highly abstract features of the target customer's behavior and represents complex patterns in multi-dimensional data. This vector serves as an intermediate result of the feature extraction network and will be used in the subsequent decoding stage.

[0082] In the decoding process, the first deconvolution layer receives the 256-dimensional feature vector output by the encoding layer and performs deconvolution using 128 3×3 convolution kernels. This step aims to gradually restore the spatial dimensions of the data, generate 128 feature maps, and reconstruct the structural information of some of the original behavior data.

[0083] Subsequently, the first upsampling layer receives the output of the first deconvolution layer and further restores the spatial resolution of the data by performing a 2×2 upsampling operation to generate 128 enlarged feature maps. This operation can expand the size of the feature map and gradually approach the spatial structure of the original input data.

[0084] Next, the second deconvolution layer receives the output of the first upsampling layer and performs further deconvolution using 64 3×3 convolution kernels. This operation aims to restore higher spatial resolution and generate 64 feature maps, which are similar to the output of the first convolution layer, but have been processed in multiple layers to extract deeper features.

[0085] The second upsampling layer receives the output of the second deconvolution layer and performs a 2×2 upsampling operation to restore the feature map to a spatial resolution similar to that of the first convolution layer. At this point, the spatial structure of the data has been basically restored and is ready to enter the final output layer.

[0086] Finally, the first output layer receives the output of the second upsampling layer and further processes the data through an additional convolutional layer to generate reconstructed data similar to the first input layer. Compared with the original input data, these reconstructed data retain key information while removing unnecessary noise. In addition, the first output layer also generates a 256-dimensional high-dimensional nonlinear behavioral feature vector, which represents the abstract behavioral characteristics of the target customer. This vector will be used in the subsequent credit assessment process. Through this combined operation of convolution and deconvolution, the system can effectively extract and restore the multi-dimensional behavioral characteristics of the target customer, providing a solid data foundation for subsequent assessment tasks.

[0087] Furthermore, the anomaly detection network adopts a recursive neural network based on a multi-head self-attention mechanism, including: The second input layer is used to receive the nonlinear behavior feature vector output from the feature extraction network; the spatiotemporal graph convolution layer is used to receive the nonlinear behavior feature vector provided by the second input layer, and the spatiotemporal graph convolution layer constructs a spatiotemporal graph based on the graph convolution network and the time dimension, and is specifically used for: By arranging the target customers' multi-dimensional behavior data and multi-dimensional behavior characteristics in chronological order, a spatiotemporal graph is formed, in which nodes represent the behavior states at different time points, and edges represent the behavior relationships between different time points; The graph convolutional network is used to convolve the nodes in the spatiotemporal graph to capture the dependency of user behavior in the time dimension and the spatial characteristics of behavior data, and obtain a 128-dimensional feature vector after spatiotemporal convolution. The multi-head self-attention layer is used to receive the 128-dimensional feature vector output from the spatiotemporal graph convolution layer, and use the multi-head self-attention mechanism to perform weighted processing on the key behavior patterns to generate a 1024-dimensional feature vector; The recursive neural network layer receives the 1024-dimensional feature vector from the multi-head self-attention layer, combines it with the long short-term memory network to process the time series characteristics of the behavior data, and outputs a 512-dimensional time series feature vector to represent the time characteristics of the abnormal behavior; The second output layer is used to receive the 512-dimensional time series feature vector provided by the recurrent neural network layer to generate the abnormal behavior prediction probability and abnormal behavior detection results of the target customers.

[0088] The anomaly detection network uses a recursive neural network based on a multi-head self-attention mechanism to identify and predict abnormal behaviors of target customers. The network consists of multiple layers, which gradually processes nonlinear behavioral feature vectors from the feature extraction network and finally generates the predicted probability and detection results of abnormal behaviors.

[0089] First, the second input layer of the anomaly detection network receives nonlinear behavioral feature vectors from the feature extraction network. These feature vectors are highly abstract behavioral representations extracted from the multi-dimensional behavioral data of target customers, including financial behavior, consumption behavior, social behavior, etc. These feature vectors capture the complexity and potential patterns of customer behavior as input to subsequent layers.

[0090] Next, the spatiotemporal graph convolution layer receives the nonlinear behavior feature vector and further analyzes the dependencies of customer behaviors in time and space dimensions by constructing a spatiotemporal graph. The nodes of the spatiotemporal graph represent the customer behavior states at different time points, and the edges represent the temporal correlations of these behaviors. By arranging the multi-dimensional behavior data of target customers in chronological order, the spatiotemporal graph can effectively capture the temporal changes and spatial characteristics of behaviors. On this basis, the graph convolutional network performs convolution operations on the nodes in the spatiotemporal graph. The convolution process can extract the relationships and patterns between nodes and generate a 128-dimensional feature vector. These feature vectors not only contain the changes in behaviors at different time points, but also reflect the temporal dependencies between behaviors, which can effectively describe the dynamic changes in customer behaviors.

[0091] After the spatiotemporal graph convolution layer, the multi-head self-attention layer receives a 128-dimensional feature vector and performs weighted processing on key behavior patterns through the self-attention mechanism. The core idea of ​​the multi-head self-attention mechanism is to process data in parallel on multiple "heads" to enhance the network's ability to focus on different feature dimensions. Each "head" independently calculates the attention weight to identify key behavior patterns in the input feature vector. Through this mechanism, the network can more accurately identify possible anomalies in customer behavior. The output results of each head will be spliced ​​to generate a 1024-dimensional feature vector, which represents the important patterns and anomaly possibilities in customer behavior.

[0092] Next, the recurrent neural network layer receives the 1024-dimensional feature vector from the multi-head self-attention layer and further processes the time series characteristics of this data through the long short-term memory network. LSTM can capture changes in behavior over a long time span and handle short-term and long-term behavioral dependencies. The recurrent neural network outputs a 512-dimensional time series feature vector, which represents the characteristics of the customer's abnormal behavior in the time dimension. These feature vectors not only take into account the current behavior state, but also combine the customer's past behavior sequence, thereby improving the timeliness and accuracy of abnormal behavior detection.

[0093] Finally, the second output layer receives the 512-dimensional time series feature vector from the recurrent neural network layer and generates the target customer's abnormal behavior prediction probability and detection results. The prediction probability indicates the possibility of abnormal behavior of the customer in a certain period of time in the future, and the detection result determines whether the customer has already committed abnormal behavior based on historical behavior data. These output data will be used for subsequent credit risk assessment and decision-making, providing an important reference for credit approval.

[0094] Furthermore, the multi-layer fusion network is implemented by a dynamic routing capsule network, and the dynamic routing capsule network includes: The third input layer is used to receive the nonlinear behavior feature vector provided by the feature extraction network, the abnormal behavior detection results and abnormal behavior prediction probability provided by the anomaly detection network, the credit report data, and the economic indicator data; the input data of the third input layer is integrated into a 919-dimensional feature vector as the input of the subsequent capsule layer; The first capsule layer is used to receive the 919-dimensional feature vector of the third input layer, and process different types of input features through multiple capsule units, wherein each capsule unit contains 8 neurons, which are responsible for processing data of different dimensions; the output of each capsule unit is an 8-dimensional vector, which represents the feature information of the input feature in different dimensions; and after passing through the first capsule layer, the output is a 64-dimensional feature vector; the dynamic routing mechanism of the first capsule layer includes: dynamically calculating the information transmission weight between capsule units according to the output of the capsule unit; the vector length of the capsule unit output represents the feature strength, and dynamically adjusting the information transmission path between different capsule units to strengthen the capsule output related to the key features; The second capsule layer is used to receive the 64-dimensional vector from the first capsule layer, and further fuse and process feature information from different sources; the second capsule layer includes multiple first capsule units, wherein each first capsule unit includes 16 neurons, and is used to process multi-dimensional features; the second capsule layer divides the 64-dimensional feature vector and inputs it into each capsule unit through a dynamic routing mechanism; the output of the second capsule layer is a 128-dimensional feature vector, which reflects the deep correlation between the target customer behavior data, economic data and anomaly detection results; A feature fusion layer is used to receive the 128-dimensional feature vector output by the second capsule layer and perform feature fusion through a fully connected layer; the output of the feature fusion layer is a 64-dimensional fused feature vector; The third output layer is used to receive the 64-dimensional fused feature vector from the feature fusion layer and generate a comprehensive credit risk assessment result of the target customer. The third output layer is specifically used to: Through the third fully connected layer, the 64-dimensional feature vector is mapped into a credit risk score, which represents the credit risk level of the target customer; The fourth fully connected layer is used to identify and output key factors that affect the customer's credit risk, and the output is a set of classification labels or specific values; wherein the key factors include abnormal behavior, credit history, and changes in the economic environment; Through the time series analysis layer, the 64-dimensional feature vector is combined with historical behavior data for prediction to generate the credit risk change trend for a specified period in the future, and the output is a future risk score that represents the customer's potential credit risk in the future.

[0095] First, the third input layer receives feature data from multiple sources, including nonlinear behavior feature vectors provided by the feature extraction network, abnormal behavior detection results and predicted probabilities provided by the anomaly detection network, and the customer's credit report data and macroeconomic indicators. These data are integrated into a 919-dimensional feature vector, which combines the customer's behavioral characteristics, historical credit information, and external economic environment as the basis for subsequent processing. The input data is preprocessed to ensure that data of different dimensions have consistent formats and comparability.

[0096] Next, the first capsule layer receives the 919-dimensional feature vector from the third input layer and uses multiple capsule units to process different types of input features in parallel. Each capsule unit contains 8 neurons, which are responsible for processing specific feature dimensions. The output of the capsule unit is an 8-dimensional vector, which represents the feature information of the input features in different dimensions. This vectorized output can capture the key information in the input data and optimize the information transmission path between capsule units through a dynamic routing mechanism. The core of the dynamic routing mechanism is to dynamically adjust the weights and connection paths between different capsule units according to the length of the output vector of the capsule unit to strengthen the output related to the key features. This mechanism ensures that the system can focus on the most important feature information and reduce the interference of noise and irrelevant information. After this layer of processing, the first capsule layer outputs a 64-dimensional feature vector, which represents the preliminary extracted feature information.

[0097] Subsequently, the second capsule layer further receives the 64-dimensional vector from the first capsule layer, and continues to process and fuse feature information from multiple data sources. The second capsule layer contains multiple capsule units, each of which contains 16 neurons to process more complex multi-dimensional features. This layer divides the 64-dimensional feature vector and inputs it into each capsule unit through a further dynamic routing mechanism. Through this process, the capsule network is able to capture deep associations between different features, such as the relationship between customer behavior patterns and macroeconomic data. Finally, the second capsule layer outputs a 128-dimensional feature vector, which combines customer behavior data, economic data, and anomaly detection results, and reveals the complex interactions between dimensions.

[0098] The feature fusion layer receives the 128-dimensional feature vector from the second capsule layer and fuses these features through a fully connected layer. The fully connected layer unifies the feature information from different sources through weighted processing to ensure that the system can comprehensively consider the influencing factors of each dimension. After this layer of processing, the output is a 64-dimensional fused feature vector, which represents the comprehensive credit risk characteristics of the customer.

[0099] Finally, the third output layer receives the 64-dimensional fused feature vector from the feature fusion layer and generates a comprehensive credit risk assessment result for the customer. This layer achieves the final assessment output through multiple fully connected layers. First, the third fully connected layer maps the 64-dimensional feature vector to a credit risk score, which represents the customer's credit risk level and is a quantitative result of the customer's overall credit status. Secondly, the fourth fully connected layer is responsible for identifying and outputting the key factors that affect the customer's credit risk. These key factors can include abnormal behavior, negative events in historical credit records, and changes in the macroeconomic environment. The system outputs these key factors in the form of classification labels or numerical values ​​to help credit assessors identify the main sources of credit risk for customers.

[0100] In addition, the system also predicts the future trend of the customer's credit risk through the time series analysis layer. The time series analysis layer combines the customer's historical behavior data and current feature vectors to predict the customer's credit risk change trend in the future (such as 6 months or 1 year). The system ultimately outputs a future risk score, which represents the customer's potential credit risk level in the future. This prediction can help credit institutions prevent possible credit risks in advance and make more accurate credit decisions.

[0101] The following is the reference implementation code of the neural network model.

[0102] import torch; import torch.nn as nn; import torch.nn.functional as F; # Feature extraction network: multi-layer convolutional autoencoder class FeatureExtractionNetwork(nn.Module): def __init__(self): super(FeatureExtractionNetwork, self).__init__(); # The first convolutional layer receives multi-dimensional behavior data, with 64 3x3 convolution kernels; self.conv1 = nn.Conv2d(in_channels=1, out_channels=64, kernel_size=3,padding=1); # The first pooling layer, 2x2 pooling window, reduces the spatial dimension; self.pool1 = nn.MaxPool2d(kernel_size=2, stride=2); # The second convolution layer, 128 3x3 convolution kernels, extract deeper features self.conv2= nn.Conv2d(in_channels=64, out_channels=128, kernel_size=3, padding=1); # Second pooling layer; self.pool2 = nn.MaxPool2d(kernel_size=2, stride=2); # Encoding layer, compressed features are 256 dimensions; self.deconv1=nn.ConvTranspose2d(in_channels=128, out_channels=64,kernel_size=3, padding=1); self.up1 = nn.Upsample(scale_factor=2); self.deconv2=nn.ConvTranspose2d(in_channels=64, out_channels=1,kernel_size=3, padding=1); self.up2 = nn.Upsample(scale_factor=2); def forward(self, x): # First convolutional layer + ReLU activation; x = F.relu(self.conv1(x)); # The first pooling layer reduces the spatial resolution; x = self.pool1(x); # Second convolutional layer + ReLU activation; x = F.relu(self.conv2(x)); # Second pooling layer; x = self.pool2(x); # Flatten the feature map and enter the encoding layer to generate a 256-dimensional nonlinear feature vector; x = x.view(-1, 128, 8, 8); x = F.relu(self.deconv1(x)); x = self.up1(x); x = F.relu(self.deconv2(x)); x = self.up2(x); return x; # Anomaly detection network: Recurrent neural network (LSTM) based on multi-head self-attention; class AnomalyDetectionNetwork(nn.Module): def __init__(self, input_size, hidden_size, num_heads): super(AnomalyDetectionNetwork, self).__init__(); # Multi-head self-attention mechanism to process spatiotemporal data; self.multihead_attention=nn.MultiheadAttention(embed_dim=input_size,num_heads=num_heads); # LSTM layer, used to process time series features; self.lstm=nn.LSTM(input_size=input_size,hidden_size=hidden_size, num_layers=2, batch_first=True); # The fully connected layer outputs the detection results and predicted probabilities of abnormal behaviors; self.fc_out = nn.Linear(hidden_size, 2); def forward(self, x): # Multi-head self-attention, processing feature vectors; attn_output, _ = self.multihead_attention(x, x, x); # LSTM processes time series features; lstm_out, (hn, cn) = self.lstm(attn_output); # Get the LSTM output at the last moment; lstm_out = lstm_out[:, -1, :]; # Output abnormal behavior detection results and predicted probabilities; output = self.fc_out(lstm_out); return output; # Multi-layer fusion network: capsule network implementation; class CapsuleLayer(nn.Module): def__init__(self,num_capsules, num_routes, in_dim, out_dim): super(CapsuleLayer, self).__init__(); self.num_capsules = num_capsules; self.num_routes = num_routes; self.in_dim = in_dim; self.out_dim = out_dim; # Linear layer for each capsule unit; self.capsules = nn.ModuleList([nn.Linear(in_dim, out_dim); for _ inrange(num_capsules)]); def forward(self, x): # Calculate the output of each capsule; u_hat = [capsule(x) for capsule in self.capsules]; u_hat = torch.stack(u_hat, dim=1)# Combine all capsule units; b_ij=torch.zeros(x.size(0),self.num_capsules, self.num_routes).to(x.device); for i in range(3):# Dynamic routing iteration times; c_ij = F.softmax(b_ij, dim=2)# Calculate weights; v_j = self.squash(s_j) # Squeeze nonlinear activation; s_j_norm = torch.norm(s_j, dim=-1, keepdim=True); class CapsuleNetwork(nn.Module): def __init__(self): super(CapsuleNetwork, self).__init__(); # The third input layer integrates nonlinear feature vectors, anomaly detection results and economic data; self.fc1 = nn.Linear(919, 128); # The first capsule layer, 64 capsule units, each capsule has 8-dimensional output; self.capsule1=CapsuleLayer(num_capsules=64, num_routes=128, in_dim=128, out_dim=8); # The second capsule layer has 128 capsule units and each capsule has 16-dimensional output; self.capsule2=CapsuleLayer(num_capsules=128, num_routes=64, in_dim=8,out_dim=16); #Fully connected layer, used for feature fusion; self.fc3 = nn.Linear(64, 1) # Credit risk score; self.fc4 = nn.Linear(64, 3)# Key risk factors (3 factors); def forward(self, x): # Initial processing, nonlinear feature vector integration; x = F.relu(self.fc1(x)); # First capsule layer processing; capsule1_out = self.capsule1(x); # Second capsule layer processing; capsule2_out = self.capsule2(capsule1_out); # Feature fusion layer; x = F.relu(self.fc2(capsule2_out.view(x.size(0), -1))); # Output credit risk score; credit_score = self.fc3(x); # Output key risk factors; risk_factors = self.fc4(x) return credit_score, risk_factors; # Comprehensive neural network model, combining various sub-networks together; class CreditRiskAssessmentModel(nn.Module): def __init__(self): super(CreditRiskAssessmentModel, self).__init__(); # Define each sub-network; self.feature_extractor = FeatureExtractionNetwork(); self.anomaly_detector= AnomalyDetectionNetwork(input_size=256,hidden_size=512,num_heads=8); self.capsule_network = CapsuleNetwork(); def forward(self, behavior_data, credit_report_data, economic_data): # Feature extraction network, extracting behavioral features; behavior_features = self.feature_extractor(behavior_data); # Anomaly detection network, detects abnormal behavior and predicts future abnormal behavior anomaly_detection_output= self.anomaly_detector(behavior_features); # Integrate behavioral characteristics, anomaly detection results, credit reports and economic data; combined_input=torch.cat([behavior_features, anomaly_detection_output, credit_report_data, economic_data], dim=1); credit_score,risk_factors=self.capsule_network(combined_input); return credit_score, risk_factors; In order to train this neural network model, we first need to prepare a training set containing multi-dimensional behavioral data, credit report data, and economic indicator data. Each data sample should include the target customer's behavioral data, credit information, and corresponding credit risk labels. The goal of the training process is to optimize the parameters of the network so that the model can accurately predict credit risk scores and effectively detect abnormal behavior.

[0103] 1. Data preparation: Divide the training data into input data and corresponding labels (i.e., the customer's credit risk score). The input data includes the target customer's multi-dimensional behavior data, abnormal behavior information, credit report data, and economic indicator data. It is necessary to ensure that the data is properly preprocessed, such as normalization or standardization, so that the network can learn more effectively.

[0104] 2. Initialize the model: Create a model instance, including a feature extraction network, anomaly detection network, and a multi-layer fusion network. These sub-networks are combined into a complete model that can extract features from the input multi-dimensional data, detect anomalies, and generate credit risk assessment results.

[0105] 3. Define the loss function and optimizer: Select the appropriate loss function and optimizer. For example, you can use the mean square error (MSE) as the loss function for credit score prediction, or the cross entropy loss function (Cross Entropy Loss) for the classification of abnormal behavior. The optimizer can choose the common Adam optimizer, which can quickly adjust the model parameters.

[0106] 4. Model training: Input data and labels are fed into the model for training. In each training iteration, the model calculates the prediction and compares it with the true label. Through backpropagation, the optimizer updates the weights in the model so that the prediction gradually approaches the true value. During this process, the loss value should gradually decrease, indicating that the model's predictive ability is gradually improving.

[0107] 5. Model evaluation: During the training process, use the validation data set to regularly evaluate the model performance. The validation data set is not used in the training and is only used to evaluate the performance of the model on unseen data. Based on the results of the validation set, the model hyperparameters, optimizer settings, etc. can be adjusted to improve the generalization ability of the model.

[0108] Step S104: Generate credit approval recommendations and personalized loan plans based on the credit risk assessment results output by the neural network model.

[0109] Step S104 aims to provide target customers with credit approval recommendations and personalized loan plans based on the credit risk assessment results generated by the neural network model. Specifically, this step will provide support for credit decision-making by conducting a comprehensive analysis of the customer's credit risk score, the identification results of key risk factors, and future credit risk predictions.

[0110] First, the system will use the credit risk score generated in step S103 as the core basis, and the score value is usually in the range of 0 to 1. The score directly reflects the overall credit health of the customer. The higher the value, the lower the credit risk of the customer. For customers with high credit scores, the system will recommend higher credit limits and more relaxed loan terms. For customers with lower credit scores, the system will recommend more conservative credit decisions, such as reducing loan limits or tightening loan conditions.

[0111] At the same time, the system will conduct an in-depth analysis of the identification results of key risk factors. By identifying the specific factors behind the customer's credit risk, such as a high frequency of abnormal behavior, an unstable economic environment, and overdue credit records, the system can develop more targeted loan plans. For example, if the identified major risk factors are related to fluctuations in the customer's financial behavior, the system may recommend shortening the loan term or increasing the loan interest rate to reduce potential risks. In addition, the system can manage customers in layers according to different risk factors and develop differentiated credit policies for customers at different risk levels to balance loan risks and benefits.

[0112] Future credit risk prediction is also an important reference in the decision-making process. The system will combine the credit risk prediction results for a period of time in the future (such as 6 months or 1 year) to evaluate whether the customer's credit risk during the loan period has an upward trend. If the model predicts that the customer's future credit risk will increase significantly, the system may recommend a more conservative credit decision, such as shortening the loan period or setting up a reassessment mechanism in advance. For customers whose credit risk is relatively stable or declining in the future, the system will recommend more flexible loan terms, appropriately extend the repayment period or lower the interest rate to attract high-quality customers.

[0113] When generating personalized loan plans, the system is not only based on credit risk assessment results, but also combines the customer's personalized needs and preferences, such as repayment ability, loan purpose, and the repayment method chosen by the customer. By analyzing the customer's historical behavior and preferences, the system can recommend loan products that meet their needs and improve customer satisfaction. For example, for customers with stable repayment habits, the system may recommend a plan with a longer repayment period and a low interest rate, while for customers who prefer quick repayment, it will recommend a short-term, high-interest loan product.

[0114] Finally, the system integrates all the analysis results to generate credit approval recommendations and personalized loan plans. This result will be submitted to the credit approval department as a basis for decision-making, and customers will also receive personalized loan plan recommendations to help them make the most suitable choice.

[0115] In the above-mentioned embodiment, a credit assessment method based on artificial intelligence and big data technology is provided. Correspondingly, the present application also provides a credit assessment device based on artificial intelligence and big data technology. Since this embodiment, i.e., the second embodiment, is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The embodiment described below is only illustrative.

[0116] The second embodiment of the present application provides a credit assessment device based on artificial intelligence and big data technology, including: A collection unit, used to collect multi-dimensional behavior data of target customers, wherein the multi-dimensional behavior data includes financial behavior data, social behavior data, consumption behavior data, geographic location behavior data, and network behavior data; An obtaining unit is used to obtain the multi-dimensional behavior characteristics of the target customer according to the multi-dimensional behavior data, wherein the multi-dimensional behavior characteristics include the behavior stability index, abnormal behavior frequency, behavior trend data and multi-dimensional behavior correlation data of the target customer; the behavior stability index is used to evaluate the consistency and volatility of the target customer in each behavior dimension; the abnormal behavior frequency refers to the frequency of abnormal events that deviate from the normal state in the behavior pattern of the target customer; the behavior trend data is used to reflect the change trend of the target customer's behavior over time and its development speed; the multi-dimensional behavior correlation data is used to reflect the mutual correlation between the behavior data of different dimensions; A generating unit, for generating a credit risk assessment result based on a pre-trained neural network model and according to the target customer's credit report data, environmental risk assessment data, multi-dimensional behavior data and multi-dimensional behavior characteristics; wherein the credit risk assessment result includes a credit risk score, identification results of key risk factors and future credit risk forecasts; the environmental risk assessment data includes economic indicator data, and the economic indicator data includes GDP growth rate, unemployment rate and industry risk data of the industry in which the target customer is engaged; The approval unit is used to generate credit approval suggestions and personalized loan plans based on the credit risk assessment results output by the neural network model.

[0117] Although the present application is disclosed as above in the form of a preferred embodiment, it is not intended to limit the present application. Any person skilled in the art may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined in the present application.

Claims

1. A credit assessment method based on artificial intelligence and big data technology, characterized in that: include: Collect multi-dimensional behavior data of target customers, including financial behavior data, social behavior data, consumption behavior data, geographic location behavior data, and network behavior data; According to the multi-dimensional behavior data, the multi-dimensional behavior characteristics of the target customer are obtained, wherein the multi-dimensional behavior characteristics include the behavior stability index, abnormal behavior frequency, behavior trend data and multi-dimensional behavior correlation data of the target customer; the behavior stability index is used to evaluate the consistency and volatility of the target customer in each behavior dimension; the abnormal behavior frequency refers to the frequency of abnormal events that deviate from the normal state in the behavior pattern of the target customer; the behavior trend data is used to reflect the change trend of the target customer's behavior over time and its development speed; the multi-dimensional behavior correlation data is used to reflect the mutual correlation between the behavior data of different dimensions; Generate credit risk assessment results based on the pre-trained neural network model and the target customer's credit report data, environmental risk assessment data, multi-dimensional behavior data and multi-dimensional behavior characteristics; wherein the credit risk assessment results include credit risk scores, identification results of key risk factors and future credit risk forecasts; the environmental risk assessment data include economic indicator data, and the economic indicator data include GDP growth rate, unemployment rate and industry risk data of the industry in which the target customer is engaged; Based on the credit risk assessment results output by the neural network model, credit approval recommendations and personalized loan plans are generated.

2. The credit assessment method according to claim 1, characterized in that: The neural network model includes a feature extraction network, an anomaly detection network and a multi-layer fusion network; The input of the feature extraction network is the multi-dimensional behavior data and multi-dimensional behavior features of the target customer; a convolutional autoencoder is used to extract features from the multi-dimensional behavior data of the target customer to obtain a nonlinear behavior feature vector; The input of the anomaly detection network is the nonlinear behavior feature vector provided by the feature extraction network; the anomaly detection network is implemented by a recursive neural network based on a multi-head self-attention mechanism to obtain abnormal behavior detection results and prediction probabilities of abnormal behaviors; The input of the multi-layer fusion network is the nonlinear behavior feature vector provided by the feature extraction network, the abnormal behavior detection results and the predicted probability of abnormal behavior provided by the anomaly detection network, the credit report data and the economic indicator data; the multi-layer fusion network is implemented using a capsule network to obtain the credit risk assessment results of the target customers.

3. The credit assessment method according to claim 2, characterized in that: The multi-layer convolutional autoencoder used in the feature extraction network includes: The first input layer is used to receive multi-dimensional behavior data and multi-dimensional behavior characteristics of target customers; A first convolutional layer, used to receive the multi-dimensional behavior data provided by the first input layer, wherein the first convolutional layer uses 64 3×3 convolution kernels to perform a convolution operation to extract local features, and outputs 64 feature maps; The first pooling layer is used to receive the output of the first convolutional layer, use a 2×2 pooling window to perform a maximum pooling operation, reduce the spatial dimension of the feature map, and output 64 pooled feature maps; The second convolution layer is used to receive the output of the first pooling layer. The second convolution layer uses 128 3×3 convolution kernels to extract deeper local features, capture temporal dependencies and nonlinear relationships, and outputs 128 feature maps. The second pooling layer is used to receive the output of the second convolutional layer. The second pooling layer uses a 2×2 pooling window to perform a maximum pooling operation to further reduce the spatial dimension of the features, and the output is 128 pooled feature maps; The encoding layer receives the output of the second pooling layer and compresses the feature vector into a 256-dimensional high-dimensional nonlinear behavior feature vector through a fully connected layer as the intermediate result of the feature extraction network and used in the decoding stage. A first deconvolution layer is used to receive the 256-dimensional feature vector output by the encoding layer. The first deconvolution layer uses 128 3×3 convolution kernels to perform a deconvolution operation, restore part of the spatial dimensions, and output 128 feature maps; The first upsampling layer is used to receive the output of the first deconvolution layer, perform a 2×2 upsampling operation, restore the spatial resolution of the data, and generate 128 enlarged feature maps; A second deconvolution layer is used to receive the output of the first upsampling layer. The second deconvolution layer uses 64 3×3 convolution kernels to further perform a deconvolution operation to restore a higher spatial resolution and output 64 feature maps. A second upsampling layer, used to receive the output of the second deconvolution layer, the second upsampling layer performs a 2×2 upsampling operation to restore the spatial resolution to a similar spatial resolution as the first convolution layer, and outputs 64 feature maps; The first output layer is used to further process the output of the second upsampling layer through a convolutional layer to generate reconstructed data similar to the first input layer; and to generate a 256-dimensional high-dimensional nonlinear behavioral feature vector to represent the abstract behavioral characteristics of the target customers.

4. The credit assessment method according to claim 3, characterized in that: The anomaly detection network adopts a recursive neural network based on a multi-head self-attention mechanism, including: The second input layer is used to receive the nonlinear behavior feature vector output from the feature extraction network; The spatiotemporal graph convolution layer is used to receive the nonlinear behavior feature vector provided by the second input layer. The spatiotemporal graph convolution layer constructs a spatiotemporal graph based on the graph convolution network and the time dimension, and is specifically used for: By arranging the target customers' multi-dimensional behavior data and multi-dimensional behavior characteristics in chronological order, a spatiotemporal graph is formed, in which nodes represent the behavior states at different time points, and edges represent the behavior relationships between different time points; The graph convolutional network is used to convolve the nodes in the spatiotemporal graph to capture the dependency of user behavior in the time dimension and the spatial characteristics of behavior data, and obtain a 128-dimensional feature vector after spatiotemporal convolution. The multi-head self-attention layer is used to receive the 128-dimensional feature vector output from the spatiotemporal graph convolution layer, and use the multi-head self-attention mechanism to perform weighted processing on the key behavior patterns to generate a 1024-dimensional feature vector; The recursive neural network layer receives the 1024-dimensional feature vector from the multi-head self-attention layer, combines it with the long short-term memory network to process the time series characteristics of the behavior data, and outputs a 512-dimensional time series feature vector to represent the time characteristics of the abnormal behavior; The second output layer is used to receive the 512-dimensional time series feature vector provided by the recurrent neural network layer to generate the abnormal behavior prediction probability and abnormal behavior detection results of the target customers.

5. The credit assessment method according to claim 4, characterized in that: The multi-layer fusion network is implemented by a dynamic routing capsule network, and the dynamic routing capsule network includes: The third input layer is used to receive the nonlinear behavior feature vector provided by the feature extraction network, the abnormal behavior detection results and abnormal behavior prediction probability provided by the anomaly detection network, the credit report data, and the economic indicator data; the input data of the third input layer is integrated into a 919-dimensional feature vector as the input of the subsequent capsule layer; The first capsule layer is used to receive the 919-dimensional feature vector of the third input layer, and process different types of input features through multiple capsule units, wherein each capsule unit contains 8 neurons, which are responsible for processing data of different dimensions; the output of each capsule unit is an 8-dimensional vector, which represents the feature information of the input feature in different dimensions; and after passing through the first capsule layer, the output is a 64-dimensional feature vector; the dynamic routing mechanism of the first capsule layer includes: dynamically calculating the information transmission weight between capsule units according to the output of the capsule unit; the vector length of the capsule unit output represents the feature strength, and dynamically adjusting the information transmission path between different capsule units to strengthen the capsule output related to the key features; The second capsule layer is used to receive the 64-dimensional vector from the first capsule layer, and further fuse and process feature information from different sources; the second capsule layer includes multiple first capsule units, wherein each first capsule unit includes 16 neurons, and is used to process multi-dimensional features; the second capsule layer divides the 64-dimensional feature vector and inputs it into each capsule unit through a dynamic routing mechanism; the output of the second capsule layer is a 128-dimensional feature vector, which reflects the deep correlation between the target customer behavior data, economic data and anomaly detection results; A feature fusion layer is used to receive the 128-dimensional feature vector output by the second capsule layer and perform feature fusion through a fully connected layer; the output of the feature fusion layer is a 64-dimensional fused feature vector; The third output layer is used to receive the 64-dimensional fused feature vector from the feature fusion layer and generate a comprehensive credit risk assessment result of the target customer. The third output layer is specifically used to: Through the third fully connected layer, the 64-dimensional feature vector is mapped into a credit risk score, which represents the credit risk level of the target customer; The fourth fully connected layer is used to identify and output key factors that affect the customer's credit risk, and the output is a set of classification labels or specific values; wherein the key factors include abnormal behavior, credit history, and changes in the economic environment; Through the time series analysis layer, the 64-dimensional feature vector is combined with historical behavior data for prediction to generate the credit risk change trend for a specified period in the future, and the output is a future risk score that represents the customer's potential credit risk in the future.

6. The credit assessment method according to claim 1, characterized in that: The calculation of the behavior stability index is achieved by a time-weighted average method, specifically including: Time slice the multi-dimensional behavior data and divide the behavior data into different time windows; Calculate the fluctuation range of the behavioral characteristics of each dimension within each time window; The behavioral fluctuations of each time window are weighted averaged, where more recent time windows are given greater weights, to assess the stability of the target customer’s current behavior and generate a behavioral stability index.

7. The credit assessment method according to claim 1, characterized in that: The detection of abnormal behavior frequency is achieved through threshold setting and adaptive learning mechanism based on historical behavior patterns, including: Establish a behavioral baseline pattern by analyzing the target customers’ behavioral data over the past 6 months to 1 year; Setting a threshold for deviation from historical behavior patterns, and behavioral events exceeding this threshold are marked as abnormal; By applying an adaptive learning algorithm, the threshold is updated in real time based on the new behavior data of target customers, making the model more accurate in detecting abnormal behavior.

8. The credit assessment method according to claim 1, characterized in that: The calculation of the multi-dimensional behavior correlation data adopts a method based on the Pearson correlation coefficient, specifically including: Normalize the behavioral characteristic data of target customers in different dimensions; Calculate the correlation between behavioral data of various dimensions, including financial behavior and consumer behavior, social behavior and geographic location behavior, etc.; The Pearson correlation coefficient is used to evaluate the linear correlation between behavioral data of different dimensions, generate a correlation matrix, and use it as input for subsequent credit risk assessment models.

9. The credit assessment method according to claim 1, characterized in that: The acquisition and update of the environmental risk assessment data is achieved through a dynamic macroeconomic monitoring system based on big data, specifically including: Obtain the latest GDP growth rate, unemployment rate and industry risk data from government statistics departments, industry reports and third-party economic data providers; Regularly update the economic indicators of the industries in which target customers are engaged, and dynamically adjust the credit risk of target customers in light of macroeconomic environments such as economic recession and rising industry risks.

10. A credit assessment device based on artificial intelligence and big data technology, characterized in that: include: A collection unit, used to collect multi-dimensional behavior data of target customers, wherein the multi-dimensional behavior data includes financial behavior data, social behavior data, consumption behavior data, geographic location behavior data, and network behavior data; An obtaining unit is used to obtain the multi-dimensional behavior characteristics of the target customer according to the multi-dimensional behavior data, wherein the multi-dimensional behavior characteristics include the behavior stability index, abnormal behavior frequency, behavior trend data and multi-dimensional behavior correlation data of the target customer; the behavior stability index is used to evaluate the consistency and volatility of the target customer in each behavior dimension; the abnormal behavior frequency refers to the frequency of abnormal events that deviate from the normal state in the behavior pattern of the target customer; the behavior trend data is used to reflect the change trend of the target customer's behavior over time and its development speed; the multi-dimensional behavior correlation data is used to reflect the mutual correlation between the behavior data of different dimensions; A generating unit, for generating a credit risk assessment result based on a pre-trained neural network model and according to the target customer's credit report data, environmental risk assessment data, multi-dimensional behavior data and multi-dimensional behavior characteristics; wherein the credit risk assessment result includes a credit risk score, identification results of key risk factors and future credit risk forecasts; the environmental risk assessment data includes economic indicator data, and the economic indicator data includes GDP growth rate, unemployment rate and industry risk data of the industry in which the target customer is engaged; The approval unit is used to generate credit approval suggestions and personalized loan plans based on the credit risk assessment results output by the neural network model.

Images