Data management and analysis method and system based on credit investigation information

Abstract

The invention is suitable for an information management method technology, and provides a data management analysis method and system based on credit investigation information, and the method comprises the steps: obtaining credit investigation data information; constructing edge weights according to the agent variables to form a multi-dimensional credit graph, identifying communities with similar credit behaviors according to existing data, obtaining positioning similar crowds, and judging contribution degree scores according to contribution degrees of white users; constructing a multi-dimensional credit behavior trajectory diagram according to the real-time behavior data flow and the platform data, identifying abrupt change nodes in the trajectory diagram, and judging whether the behavior change forms a signal of credit risk offset according to the abrupt change nodes; and quantifying the positive or negative influence amplitude of each behavior feature on the final score according to the final score, obtaining updated evidence material information according to the score elements, and obtaining a new score suggestion value according to the evidence material information. The key problems that in a traditional credit investigation system, white users are difficult to model, the scoring result is static and rigid, and appeal cannot be fed back are effectively solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of information management methods, and in particular relates to a method and system for managing and analyzing credit information data. Background Art

[0002] Currently, credit scoring systems are widely used in multiple scenarios, including financial credit, e-commerce credit, and leasing services. Their core purpose is to quantitatively assess users' credit behavior to assist with risk control and decision-making. However, existing credit scoring systems primarily rely on users' credit history data, such as loan records, repayment behavior, and credit card usage. These systems are unable to effectively cover users with no credit history or minimal credit activity, resulting in a large number of potentially low-risk users being excluded from traditional financial services. Furthermore, traditional scoring models generally employ static rules or fixed weight calculations, making it difficult to dynamically reflect real-time changes in user behavior. They lack the ability to promptly identify and respond to credit risks caused by emergencies, which can easily lead to delayed or distorted scoring results and reduce the sensitivity and accuracy of risk control.

[0003] On the other hand, most current credit scoring systems are closed models, lacking transparency in the scoring process. Users typically only receive the final score, lacking visibility into the score's composition and the specific impact of various behavioral factors. This not only undermines user trust in the scoring results, but also fails to meet user feedback and appeal needs regarding scoring disputes. In the event of a misjudgment or abnormal deduction, users are often unable to provide valid evidence for score revision, impacting their legitimate rights and interests in financial activities.

[0004] Therefore, there is an urgent need for an intelligent scoring method that can integrate multi-source data, is applicable to white-household users, supports dynamic adjustment of behavioral changes, and has an explainability and appeal mechanism, so as to improve the applicability, real-time response capability and user participation of the scoring model, and promote the development of the credit reporting system in a more open, fair and trustworthy direction. Summary of the Invention

[0005] The purpose of the embodiment of the present invention is to provide a credit information data management and analysis method, which aims to solve the problems raised in the third part of the background technology.

[0006] The embodiment of the present invention is implemented as follows: a method for managing and analyzing credit information data, the method comprising: Obtaining credit information data, where the credit information is obtained from multiple heterogeneous sources, including bank, consumer finance, and operator data; Based on the proxy variables, edge weights are constructed to form a multi-dimensional credit map. Based on the existing data, communities with similar credit behaviors are identified to locate similar groups of people. The contribution scores are determined based on the contribution of white household users. Build a multi-dimensional credit behavior trajectory diagram based on real-time behavioral data streams and platform data, identify mutation nodes in the trajectory diagram, and determine whether the behavioral change constitutes a signal of credit risk deviation based on the mutation nodes; The positive or negative impact of each behavioral characteristic on the final score is quantified based on the final score, updated evidence material information is obtained based on the scoring factors, and a new score recommendation value is obtained based on the evidence material information.

[0007] Preferably, the steps of constructing edge weights based on proxy variables to form a multidimensional credit graph, identifying communities with similar credit behaviors based on existing data, locating similar groups of people, and determining contribution scores based on the contribution of white household users specifically include: Determine white household users based on credit data. White household users are users with no credit history. Proxy variables are obtained based on white household users. The proxy variables include utility payment records such as water, electricity, and gas, and fulfillment rates and return rates on e-commerce platforms. Determine financial behavior capabilities based on proxy variables, construct edge weights based on proxy variables to form a multidimensional credit map, obtain existing data, identify communities with similar credit behaviors based on existing data, and locate similar groups of people; The contribution of existing white-household users is obtained, where the contribution includes continuous payment, e-commerce positive reviews, and unstable location, and a contribution score is determined based on the contribution of the white-household users.

[0008] Preferably, the steps of constructing a multi-dimensional credit behavior trajectory diagram based on the real-time behavior data stream and platform data, identifying mutation nodes in the trajectory diagram, and determining whether the behavior change constitutes a signal of credit risk deviation based on the mutation nodes specifically include: Obtain real-time behavioral data streams, including financial, e-commerce, and public service data, and connect to the platform through authorized API interfaces to obtain platform data; Construct a multi-dimensional credit behavior trajectory map based on real-time behavior data streams and platform data. The dimensions of the trajectory map represent key credit factors, and identify mutation nodes in the trajectory map. Based on the mutation node, it is determined whether the behavioral change constitutes a signal of credit risk deviation. If it is determined that there is a credit risk deviation, the score is lowered by calculating the correction factor, and the node record is retained for trend monitoring.

[0009] Preferably, the steps of quantifying the positive or negative impact of each behavioral feature on the final score based on the final score, obtaining updated evidence material information based on the scoring factors, and obtaining a new score recommendation value based on the evidence material information specifically include: Obtain the revised final score, quantify the positive or negative impact of each behavioral characteristic on the final score based on the final score, and visualize the quantified results; Obtaining scoring components, including credit enhancement factors, credit risk factors, and neutral characteristics, and obtaining updated evidentiary information based on the scoring components, which is used for scoring appeals; A new score recommendation value is obtained based on the evidence material information. If the recommendation value is significantly different from the original score and the evidence material information is credible, the score is revised and the score revision path is recorded.

[0010] Preferably, the platform includes bank credit investigation, payment platform, e-commerce platform and operator.

[0011] Another object of an embodiment of the present invention is to provide a credit information data management and analysis system, the system comprising: A credit information module, which obtains credit data information from multiple heterogeneous sources, including bank, consumer finance, and operator data; The contribution scoring module constructs edge weights based on proxy variables to form a multi-dimensional credit map. Based on existing data, it identifies communities with similar credit behaviors, locates similar groups of people, and determines the contribution score based on the contribution of white household users. The risk identification module constructs a multi-dimensional credit behavior trajectory diagram based on real-time behavioral data streams and platform data, identifies mutation nodes in the trajectory diagram, and determines whether the behavioral change constitutes a signal of credit risk deviation based on the mutation nodes; The scoring interpretation module quantifies the positive or negative impact of each behavioral feature on the final score based on the final score, obtains updated evidence material information based on the scoring factors, and obtains a new score recommendation value based on the evidence material information.

[0012] Preferably, the contribution scoring module includes: An agent variable unit determines white-account users based on credit data. The white-account users are users with no credit history, and obtains agent variables based on the white-account users. The agent variables include utility payment records such as water, electricity, and gas, and fulfillment rates and return rates on e-commerce platforms. The credit graph unit determines financial behavior capabilities based on proxy variables, constructs edge weights based on proxy variables to form a multi-dimensional credit graph, obtains existing data, identifies communities with similar credit behaviors based on the existing data, and locates similar groups of people; The contribution scoring unit obtains the contribution of existing white-household users, where the contribution includes continuous payment, e-commerce praise and unstable location, and determines the contribution score based on the contribution of white-household users.

[0013] Preferably, the risk identification module includes: The real-time behavior unit obtains real-time behavior data streams, including financial, e-commerce, and public service data, and obtains platform data through the authorized API interface connection platform; The behavior trajectory graph unit constructs a multi-dimensional credit behavior trajectory graph based on real-time behavior data streams and platform data. The trajectory graph dimensions represent key credit factors and identify mutation nodes in the trajectory graph. The risk shift unit determines whether the behavioral change constitutes a signal of credit risk shift based on the mutation node. If it is determined that there is a credit risk shift, the score is lowered by calculating the correction factor and the node record is retained for trend monitoring.

[0014] Preferably, the score explanation module includes: The score interpretation unit obtains the revised final score, quantifies the positive or negative impact of each behavioral characteristic on the final score based on the final score, and visualizes the quantified results; A material information updating unit, which obtains scoring components, including credit enhancement factors, credit risk factors, and neutral characteristics, and obtains updated evidence material information based on the scoring components, which is used for scoring appeals; The scoring correction unit obtains a new scoring recommendation value based on the evidence material information. If the recommendation value is significantly different from the original score and the evidence material information is credible, the score is corrected and the scoring correction path is recorded.

[0015] Preferably, the platform includes bank credit investigation, payment platform, e-commerce platform and operator.

[0016] The embodiment of the present invention provides a method for managing and analyzing credit information data, which provides an intelligent credit scoring method that integrates multi-source credit data fusion, proxy variable modeling, graph structure community identification, dynamic analysis of behavioral trajectories, and score correction. It effectively solves key problems in traditional credit reporting systems, such as the difficulty in modeling undocumented users, static and rigid scoring results, and the inability to feedback complaints. By introducing proxy variables such as public payment, e-commerce fulfillment, communication behavior, and location information, a multi-dimensional feature vector of users is constructed, and similar groups in the credit behavior graph are identified in a graph structure manner, thereby achieving behavioral classification and initial score estimation for undocumented users. This mechanism significantly expands the scope of application of the scoring system in user groups without credit data, and improves the coverage and early judgment capabilities of the scoring model.

[0017] During the scoring process, this method constructs a multidimensional credit behavior trajectory map by acquiring real-time behavioral data streams from platforms such as finance, e-commerce, and public services. It then uses a sliding window and anomaly detection algorithm to identify mutation nodes, determine whether a user exhibits credit risk deviations, and dynamically adjust the score by calculating a correction factor based on the magnitude and significance of the mutation. The final score reflects not only stable user behavior but also trends in behavioral changes.

[0018] After the scoring results are generated, the system further uses an interpretable model to quantify and visualize the contribution of each factor to the score. Users can also submit supporting evidence based on the scoring factors to appeal. After OCR and semantic verification, the evidence triggers a partial re-scoring process. If the recommended score changes significantly and the evidence is credible, the score is automatically revised and the score change path is recorded. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flowchart of a credit information data management and analysis method provided by an embodiment of the present invention; Figure 2 A flowchart of the steps of constructing edge weights based on proxy variables to form a multidimensional credit graph and determining a contribution score based on the contribution of white household users, provided in an embodiment of the present invention; Figure 3 A flowchart of the steps provided in an embodiment of the present invention for constructing a multidimensional credit behavior trajectory diagram based on real-time behavior data streams and platform data, and determining, based on mutation nodes, whether the behavior change constitutes a signal of credit risk deviation; Figure 4 A flowchart of the steps of quantifying the positive or negative impact of each behavioral characteristic on the final score based on the final score and obtaining a new score recommendation value based on the evidence material information provided in an embodiment of the present invention; Figure 5 An architecture diagram of a credit information data management and analysis system provided by an embodiment of the present invention; Figure 6 This is an architecture diagram of the contribution scoring module provided in an embodiment of the present invention; Figure 7 This is an architectural diagram of the risk identification module provided by an embodiment of the present invention; Figure 8 This is an architectural diagram of the scoring explanation module provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0021] It is understood that the terms "first," "second," etc., used herein may be used to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script, without departing from the scope of this application.

[0022] like Figure 1 As shown, a credit information data management and analysis method based on an embodiment of the present invention is provided, and the method includes: S100, obtaining credit information data, wherein the credit information data is obtained through multiple heterogeneous sources, including bank, consumer finance, and operator data.

[0023] In this step, credit information is obtained using a multi-source heterogeneous data fusion mechanism, encompassing multiple sources, including banks, consumer finance institutions, and telecom operators, to enhance the completeness and representativeness of the user's credit profile. Within the banking dimension, account transaction data, credit card usage records, repayment history, and loan contract information from traditional commercial banks and internet banks can be accessed, with particular attention paid to quantifiable indicators such as repayment frequency, proportion of bill amount, and overdue behavior. For example, if a user has multiple small loans in the past 12 months, with no overdue payments, but frequently changes loan platforms, this behavior will be included in the model as a risk boundary factor.

[0024] In the dimension of consumer finance, by accessing the data interfaces of consumer installment platforms, micro-loan systems and e-commerce financial products, we can extract users' performance in installment shopping, cash installments, BNPL (buy now, pay later) and other businesses, including approval rates, credit utilization rates and dispute records, and establish a model of users' willingness to take responsibility and financial behavior habits.

[0025] Telecom operator data serves as an auxiliary credit investigation dimension, providing users with call details, number usage duration, payment cycle patterns, roaming records and location trajectories to assess their life stability and credit consistency. For example, long-term arrears in phone bills, frequent changes of mobile phone numbers or sudden changes in call patterns will be treated as potential credit default warning factors.

[0026] S200, construct edge weights based on proxy variables to form a multi-dimensional credit map, identify communities with similar credit behaviors based on existing data, locate similar groups of people, and determine contribution scores based on the contribution of white household users.

[0027] In this step, edge weights are constructed based on proxy variables to form a multidimensional credit graph. To model the credit behavior of white-household users, edge weights are first constructed based on proxy variables, forming a multidimensional credit graph that reflects the relationships between user behavioral characteristics. Specifically, a feature vector is constructed for each user using proxy variables such as utility bill payment consistency, e-commerce compliance rate, device location stability, and phone bill payment habits. This graph structure not only preserves behavioral similarities between users but also maps implicit behavioral trend proximities.

[0028] Next, we connect white-household users to the graph and identify their top-K most similar users or communities based on the edge weights between their proxy feature vectors and those of previously labeled users. We then preliminarily estimate the white-household user's credit behavior trends across various dimensions based on the historical mean and fluctuations of user ratings within these communities. We assess the degree of fit and impact of white-household users on the characteristic centers of adjacent communities within the graph structure. If their behavior is highly consistent with the core characteristics of the community and does not significantly disrupt the structure, they are assigned a higher contribution score.

[0029] S300: Construct a multi-dimensional credit behavior trajectory diagram based on real-time behavior data streams and platform data, identify mutation nodes in the trajectory diagram, and determine whether the behavior change constitutes a signal of credit risk deviation based on the mutation nodes.

[0030] In this step, a multi-dimensional credit behavior trajectory graph is constructed based on the real-time behavior data stream and platform data. In order to achieve dynamic monitoring of user credit behavior, a multi-dimensional credit behavior trajectory graph is constructed based on the real-time behavior data stream and historical platform data. The trajectory graph is user-centric and vertically records its time series performance in each key dimension. Each dimension constitutes a behavior curve, with time as the horizontal axis and a standardized score or behavior indicator value as the vertical axis. As data continues to be input, the trajectory graph is constantly updated, which can clearly reflect the evolution trend and stability of user behavior. Through sliding windows and statistical anomaly detection methods, the system continuously monitors the curves of each dimension and captures the mutation nodes in the behavior curve, that is, the abnormal jump or interruption of the behavior indicator at a certain point in time; After identifying a sudden change, the system further assesses risk deviation based on historical behavioral baselines, dimension importance, and abnormal persistence. If the sudden change has not occurred historically, exceeds a threshold, and occurs in a highly sensitive dimension, it is identified as a credit risk deviation signal, triggering the score correction process. Following identification, the signal is marked as a high-risk deviation signal and the credit score is lowered.

[0031] S400: quantify the positive or negative impact of each behavioral feature on the final score based on the final score, obtain updated evidence material information based on the scoring factors, and obtain a new score recommendation value based on the evidence material information.

[0032] In this step, the positive or negative impact of each behavioral feature on the final score is quantified based on the final score. To address the issues of unexplainable scoring results and users' inability to provide proactive feedback, a causal analysis is performed on the behavioral features that make up the score based on the final scoring results. A marginal impact analysis is performed on each feature using local interpretable models such as SHAP values ​​or LIME to quantify its positive or negative impact on the final score.

[0033] After the scoring results are explained, if the user disagrees with a specific scoring factor, they can submit updated evidence through the front-end interactive interface, such as payment receipts, dispute revocation notices, updated professional certifications, or successful platform appeal records. Through OCR recognition and text semantic analysis, the evidence provided by the user is matched and associated with the original scoring factors, automatically screening items with corrective significance and replacing or correcting relevant feature values. After obtaining valid materials, the model reruns the scoring process, generates a new score recommendation, and compares the difference with the original score.

[0034] like Figure 2 As shown, as a preferred embodiment of the present invention, the steps of constructing edge weights based on proxy variables to form a multidimensional credit graph, identifying communities with similar credit behaviors based on existing data, locating similar groups of people, and determining contribution scores based on the contribution of white household users specifically include: S201, determining white household users based on credit data, wherein the white household users are users with no credit history, and obtaining proxy variables based on the white household users, wherein the proxy variables include utility payment records such as water, electricity, and gas, and fulfillment rates and return rates on e-commerce platforms.

[0035] In this step, white-label users are identified based on credit data. During the credit data processing process, the user's historical behavior information across banks, consumer finance, and credit platforms is analyzed to determine whether they have a complete credit record. If a user has never opened a credit card, has no loan history, and has no repayments, they are considered white-label users, meaning they have no credit history. Traditional scoring models lack core assessment criteria for white-label users, so proxy variables are constructed using alternative data to supplement their credit profiles.

[0036] Proxy variables primarily cover two areas: First, payment records for utility bills such as water, electricity, and gas, including payment continuity, any interruptions, and fluctuations in bill amounts. For example, if a user has paid their utility bills on time for 12 consecutive months, this indicates good residential stability and payment ability, positively reflecting their willingness to fulfill credit responsibilities. Second, data on e-commerce platform performance, including order completion rates, return rates, frequency of negative reviews, and post-sales dispute records. For example, if a user completed 50 e-commerce transactions in the past six months, returned only two items, and received positive reviews, this indicates stable transaction behavior and a strong willingness to fulfill contracts, which can be considered a positive proxy indicator of credit.

[0037] S202, determine financial behavior capabilities based on proxy variables, construct edge weights based on proxy variables to form a multidimensional credit map, obtain existing data, identify communities with similar credit behaviors based on the existing data, and locate similar people.

[0038] In this step, financial capacity is determined based on proxy variables. To assess a user's financial capacity in a pre-credit environment, we first quantitatively model their sense of responsibility, willingness to fulfill obligations, and lifestyle stability based on these proxy variables. These proxy variables include public payment records (such as the consistency and timeliness of utility bills), e-commerce fulfillment performance (such as order completion rates, return rates, and review content), telecommunications payment habits, and device location stability. By assigning weights and thresholds to each variable, we can initially infer whether a user possesses the financial capacity to fulfill obligations and make consistent payments on time. For example, if a user has made 12 consecutive months of payments, has an e-commerce return rate below 5%, and maintains a high and stable phone activity, they can be considered to have strong financial capacity.

[0039] Based on the aforementioned proxy variables, a multidimensional feature vector is constructed. A similarity metric (cosine similarity) is used to calculate the behavioral similarity between any two users. Users are represented as nodes in a graph structure, connected by edges, with edge weights reflecting the degree of behavioral similarity, thus forming a multidimensional credit graph. Once the graph is constructed, a graph embedding algorithm is used to map the nodes into a low-dimensional space. K-means clustering is then used to identify groups of users with similar characteristics, i.e., communities with similar credit behaviors. For example, if a majority of users in a community share the common characteristics of stable bill payments, good e-commerce performance, but no credit history, it can be inferred that the community's overall risk is low.

[0040] S203, obtaining the contribution of existing white household users, wherein the contribution includes continuous payment, e-commerce praise and unstable location, and determining the contribution score according to the contribution of the white household users.

[0041] In this step, the contribution of existing white-household users is obtained. To assess the matching quality and behavioral credibility of white-household users in the credit graph, a contribution score is calculated by calculating their performance on key proxy variable dimensions. Contribution measures the degree of consistency between white-household users' behavior and the characteristics of the center of similar communities they belong to. It mainly includes three aspects: First, payment continuity, that is, whether the user has consistently paid bills such as water, electricity, gas, and phone bills. If the user has paid bills without interruption and with low volatility for the past 12 months, their payment behavior is considered stable and has high positive contribution. Second, e-commerce positive reviews, which refers to the user's order completion rate, positive review rate, and low dispute rate on e-commerce platforms. A positive review rate above 90% indicates strong fulfillment capabilities and stable consumption behavior, which constitutes a clear positive credit signal. Third, location instability, which reflects whether the user's mobile device's geographical location frequently crosses regions or changes in short periods of time. Frequent changes of residence and unusual whereabouts are considered negative contribution factors and may indicate unstable lifestyles or secretive behavior.

[0042] When constructing the contribution score, the aforementioned proxy variables are standardized and feature weights are assigned. A matching score is calculated based on their fit within the target community's central feature vector. For example, if a user's payment continuity and positive review rate are both highly similar to those of the target group, but their positional volatility is slightly higher, a contribution score of 0.76 might be output, indicating that their overall behavior is reliable but with slight deviations. Ultimately, this contribution score is used not only to determine a user's trustworthiness within the current community but also as a key indicator for their credit score weighting.

[0043] like Figure 3 As shown in FIG. , as a preferred embodiment of the present invention, the steps of constructing a multidimensional credit behavior trajectory diagram based on the real-time behavior data stream and platform data, identifying mutation nodes in the trajectory diagram, and determining whether the behavior change constitutes a signal of credit risk deviation based on the mutation nodes specifically include: S301, obtain real-time behavior data stream, which includes financial, e-commerce and public service data, and obtain platform data through the authorized API interface docking platform.

[0044] In this step, real-time behavioral data streams are acquired. To dynamically identify and continuously update a user's credit status, a behavioral evolution sequence is constructed by acquiring real-time behavioral data streams. This data stream primarily includes three types of information: finance, e-commerce, and public services. Financial data covers account balance changes, repayment behavior, credit usage, and overdue warnings. For example, a user's failure to automatically repay a bank loan triggers a negative behavior node. E-commerce data includes order completion status, review records, return frequency, and purchase cycles. For example, three consecutive after-sales refunds are considered a potential performance deviation signal. Public service data refers to the timeliness and amount fluctuations of utility bill payments, as well as any sudden changes in urban commuting trajectories. For example, a user's phone bill payment has not been paid for two consecutive months, or the device's location signal frequently crosses provinces, all of which may indicate risk.

[0045] All data acquisition relies on standardized APIs provided by various platforms after user authorization. By establishing interfaces with banks, e-commerce platforms, and public utility service platforms, credit-related behavioral data is automatically pulled with user authorization and integrated into the user's credit history by timestamp. This process enables high-frequency collection and real-time synchronization of key credit behaviors, ensuring that the scoring model is dynamically updated based on the latest and most realistic behavioral status.

[0046] S302: Construct a multi-dimensional credit behavior trajectory diagram based on the real-time behavior data stream and platform data, wherein the dimensions of the trajectory diagram represent key credit factors, and identify mutation nodes in the trajectory diagram.

[0047] In this step, a multidimensional credit behavior trajectory diagram is constructed based on real-time behavioral data streams and platform data. Based on real-time behavioral data streams obtained from financial, e-commerce, and public service platforms, a multidimensional credit behavior trajectory diagram centered on the user is constructed. This trajectory diagram uses time as the horizontal axis and multiple key credit factors as the vertical dimensions. Each factor dimension corresponds to a time-updated behavioral sequence curve, reflecting the changing trends of users' credit behavior along different dimensions. Common dimensions include repayment history, e-commerce fulfillment, payment behavior, and location stability. For example, if a user's water and electricity payment continuity dimension has remained stable for the past six months, but two consecutive overdue payments occur in the current month, this will manifest as a significant sudden change in the trajectory of this dimension.

[0048] To identify mutation nodes in the trajectory graph, statistical anomaly detection methods such as Z-score, sliding window mean deviation, or local outlier factor (LOF) are used to dynamically monitor each behavior curve. When the behavior value at a certain point in time deviates significantly from its historical average and crosses a set threshold, it is marked as a mutation node. For example, if a user's e-commerce return rate averaged 2% over the past 12 months but suddenly increased to 12% this month, this node would be identified as a mutation and used as an early signal of potential fulfillment risk for score adjustment.

[0049] S303: Determine whether the behavior change constitutes a signal of credit risk deviation based on the mutation node. If it is determined that there is a credit risk deviation, the score is lowered by calculating the correction factor, and the node record is retained for trend monitoring.

[0050] In this step, based on the mutation node, it is determined whether the behavior change constitutes a signal of credit risk deviation. After the mutation node is identified in the multi-dimensional credit behavior trajectory diagram, it is necessary to further determine whether the behavior change constitutes a substantial credit risk deviation signal. The judgment process comprehensively considers factors such as the mutation amplitude, duration, degree of deviation from the historical baseline, and credit sensitivity of the dimension to which it belongs. If the mutation occurs in a high-weight dimension, such as repayment behavior or e-commerce fulfillment, and the amplitude of the change exceeds the statistical threshold, it means that the behavior has major abnormal characteristics. Taking the user's repayment behavior in the past 12 months as an example, if the repayment has always been on time but has been delayed twice in a row this month, and the number of delays exceeds 10 days, this behavior node will be marked as abnormal, and its deviation amplitude relative to the historical baseline will be further calculated. Combined with the importance weights of the behavioral dimensions, a correction factor is constructed.

[0051] like Figure 4 As shown, as a preferred embodiment of the present invention, the steps of quantifying the positive or negative impact of each behavioral feature on the final score according to the final score, obtaining updated evidence material information according to the scoring factors, and obtaining a new score recommendation value based on the evidence material information specifically include: S401, obtaining the corrected final score, quantifying the positive or negative impact of each behavioral feature on the final score based on the final score, and visually displaying the quantified results.

[0052] In this step, the final, revised score is obtained. After identifying mutation nodes and calculating the correction factor, the original score and the correction factor are superimposed to obtain the revised final score. This final score not only reflects the user's overall credit status but also includes dynamic adjustments for short-term risk bias. To enhance the transparency and interpretability of the scoring results, the influence of the individual behavioral characteristics that contribute to the score is further quantified. An algorithm such as the SHAP (Shapley Additive Explanations) algorithm is used to perform inverse analysis of the scoring model and calculate the marginal contribution of each characteristic to the model's predictions. Each behavioral factor is assigned a positive or negative scoring weight, indicating its contribution to or deduction from the final score. For example, a user's final score of 82 points might include a +10 contribution from utility bill consistency, a -6 contribution from occupational volatility, and a -4 contribution from e-commerce return rate. The system stores these quantitative results as scores, forming a score explanation structure.

[0053] To enhance user understanding of the scoring structure, all feature contributions are visualized, often using bar charts, score composition diagrams, or stratified score distribution charts. The charts indicate the direction and strength of each factor's impact on the score. For example, a chart might show payment behavior and e-commerce fulfillment in the positive zone, while unstable device location is in the negative zone. This helps users quickly understand the primary factors contributing to their scores and identify potential behavioral optimization areas.

[0054] S402, obtaining scoring components, including credit enhancement factors, credit risk factors, and neutral features, and obtaining updated evidence material information based on the scoring components, the evidence material information being used for scoring appeal.

[0055] In this step, the scoring components are obtained. While outputting the final score, further scoring components can be extracted, classifying all behavioral characteristics involved in the score calculation into three categories: credit enhancement factors, credit risk factors, and neutral characteristics. Credit enhancement factors are behavioral characteristics that have a significant positive impact on the score, such as consistent payment of utility bills, high e-commerce positive reviews, and a stable employment record. Credit risk factors are characteristics that negatively impact the score, such as frequent credit card returns and frequent changes in location. Neutral characteristics are variables that have a minimal impact on the score or have insignificant fluctuations, such as occasional small refunds or minor payment delays. The resulting score is presented in a structured manner, with the impact direction and weight of each factor included, to facilitate user understanding and complaint resolution.

[0056] When users have objections to the rating, they can focus on specific risk factors or key deduction items based on the above classification results and submit corresponding updated evidence materials. These materials may include payment receipts for phone bills that were mistakenly marked as arrears, records of e-commerce platform dispute cancellations, proof of position or address stability, etc. The system uses OCR and semantic analysis to identify the content of the materials and match the corresponding scoring elements, triggering a local re-scoring process after the model is verified. For example, if a user's rating is deducted due to overdue phone bills, and they provide valid payment screenshots and operator bills, the model comparison finds that it is indeed a mislabeling, and will automatically adjust the scoring factor value and update the scoring results.

[0057] S403: Obtain a new score recommendation value based on the evidence material information. If the recommendation value is significantly different from the original score and the evidence material information is credible, revise the score and record the score revision path.

[0058] In this step, a new score recommendation value is obtained based on the evidence material information. When the user submits the evidence material information related to the scoring result, the system will verify and update the corresponding scoring factor based on the material, and then re-run the scoring model to generate a new score recommendation value. The material verification process first extracts the voucher content through OCR text recognition, and then confirms its association with the specific scoring factor through semantic matching and field comparison algorithms. For example, if a user is deducted points due to overdue phone bills, the uploaded phone bill payment record is shown as automatic payment completed after identification, and the payment time is exactly the same as the bill period, the model considers the material valid and accordingly amends the value of the scoring factor from overdue to on-time payment. The scoring model recalculates the result after updating the factor. If the generated new score recommendation value is significantly different from the original score, such as exceeding the set threshold of 5 points, and the evidence material scores well in the credibility assessment, then the score recommendation will be adopted.

[0059] After the correction is completed, the scoring process will automatically record the complete path of the score adjustment, including the triggering factor, original value, modified value, material verification status, and the difference between the before and after scores, and generate a score correction log. For example, a user's initial score was 76 points, of which 6 points were mistakenly deducted due to an e-commerce dispute record due to a delayed system update. After the user submitted a revocation certificate and the model confirmed its validity, the factor was revised from dispute existing to dispute cleared, and the new score was updated to 81 points. The system archived this change as a score correction event in the user's credit history.

[0060] like Figure 5 As shown, a credit information data management and analysis system provided by an embodiment of the present invention includes: The credit information module 100 is used to obtain credit data information, which is obtained through multiple heterogeneous sources, including bank, consumer finance, and operator data.

[0061] In this system, the credit information module 100 acquires credit data using a multi-source heterogeneous data fusion mechanism, encompassing multiple sources such as banks, consumer finance institutions, and telecom operators, to enhance the completeness and representativeness of user credit profiles. Within the banking dimension, this module accesses account transaction data, credit card usage records, repayment history, and loan contract information from traditional commercial banks and internet banks, focusing specifically on quantifiable indicators such as repayment frequency, percentage of bill amount, and overdue behavior. For example, if a user has multiple small loans in the past 12 months and none are overdue, but they frequently switch loan platforms, this behavior will be modeled as a risk boundary factor.

[0062] In the dimension of consumer finance, by accessing the data interfaces of consumer installment platforms, micro-loan systems and e-commerce financial products, we can extract users' performance in installment shopping, cash installments, BNPL (buy now, pay later) and other businesses, including approval rates, credit utilization rates and dispute records, and establish a model of users' willingness to take responsibility and financial behavior habits.

[0063] Telecom operator data serves as an auxiliary credit investigation dimension, providing users with call details, number usage duration, payment cycle patterns, roaming records and location trajectories to assess their life stability and credit consistency. For example, long-term arrears in phone bills, frequent changes of mobile phone numbers or sudden changes in call patterns will be treated as potential credit default warning factors.

[0064] The contribution scoring module 200 is used to construct edge weights based on proxy variables to form a multi-dimensional credit map, identify communities with similar credit behaviors based on existing data, locate similar groups of people, and determine the contribution score based on the contribution of white household users.

[0065] In this system, the contribution scoring module 200 constructs edge weights based on proxy variables to form a multidimensional credit graph. To model the credit behavior of white-household users, edge weights are first constructed based on proxy variables, forming a multidimensional credit graph that reflects the relationships between user behavioral characteristics. Specifically, each user constructs a feature vector based on their proxy variables, such as utility bill payment continuity, e-commerce compliance rate, device location stability, and phone bill payment habits. This constructed graph structure not only preserves behavioral similarities between users but also maps the implicit proximity of behavioral trends.

[0066] Next, we connect white-household users to the graph and identify their top-K most similar users or communities based on the edge weights between their proxy feature vectors and those of previously labeled users. We then preliminarily estimate the white-household user's credit behavior trends across various dimensions based on the historical mean and fluctuations of user ratings within these communities. We assess the degree of fit and impact of white-household users on the characteristic centers of adjacent communities within the graph structure. If their behavior is highly consistent with the core characteristics of the community and does not significantly disrupt the structure, they are assigned a higher contribution score.

[0067] The risk identification module 300 is used to construct a multi-dimensional credit behavior trajectory diagram based on real-time behavior data streams and platform data, identify mutation nodes in the trajectory diagram, and determine whether the behavior change constitutes a signal of credit risk deviation based on the mutation nodes.

[0068] In this system, the risk identification module 300 constructs a multi-dimensional credit behavior trajectory diagram based on the real-time behavior data stream and platform data. In order to achieve dynamic monitoring of user credit behavior, a multi-dimensional credit behavior trajectory diagram is constructed based on the real-time behavior data stream and historical platform data. The trajectory diagram is user-centric and vertically records its time series performance in each key dimension. Each dimension constitutes a behavior curve, with the horizontal axis being time and the vertical axis being the standardized score or behavior indicator value. As data continues to be input, the trajectory diagram is constantly updated, which can clearly reflect the evolution trend and stability of user behavior. Through sliding windows and statistical anomaly detection methods, the system continuously monitors the curves of each dimension and captures the mutation nodes in the behavior curve, that is, the abnormal jump or interruption of the behavior indicator at a certain point in time; After identifying a sudden change, the system further assesses risk deviation based on historical behavioral baselines, dimension importance, and abnormal persistence. If the sudden change has not occurred historically, exceeds a threshold, and occurs in a highly sensitive dimension, it is identified as a credit risk deviation signal, triggering the score correction process. Following identification, the signal is marked as a high-risk deviation signal and the credit score is lowered.

[0069] The scoring interpretation module 400 is used to quantify the positive or negative impact of each behavioral feature on the final score based on the final score, obtain updated evidence material information based on the scoring factors, and obtain a new score recommendation value based on the evidence material information.

[0070] In this system, the score interpretation module 400 quantifies the positive or negative impact of each behavioral feature on the final score based on the final score. To solve the problems of unexplainable score results and users' inability to actively provide feedback, the score interpretation module 400 performs causal analysis on the behavioral features that constitute the score based on the final score results, and performs marginal impact analysis on each feature through local interpretable models such as SHAP value or LIME to quantify its positive or negative impact on the final score.

[0071] After the scoring results are explained, if the user disagrees with a specific scoring factor, they can submit updated evidence through the front-end interactive interface, such as payment receipts, dispute revocation notices, updated professional certifications, or successful platform appeal records. Through OCR recognition and text semantic analysis, the evidence provided by the user is matched and associated with the original scoring factors, automatically screening items with corrective significance and replacing or correcting relevant feature values. After obtaining valid materials, the model reruns the scoring process, generates a new score recommendation, and compares the difference with the original score.

[0072] like Figure 6 As shown, as a preferred embodiment of the present invention, the contribution scoring module 200 includes: The proxy variable unit 201 is used to determine white-account users based on credit data. The white-account users are users with no credit history, and obtain proxy variables based on the white-account users. The proxy variables include public payment records such as water, electricity, gas, etc. and the fulfillment rate and return rate in the e-commerce platform.

[0073] In this module, proxy variable unit 201 identifies white-account users based on credit data. During credit data processing, the user's historical behavior information across banking, consumer finance, and credit platforms is analyzed to determine whether they have a complete credit record. If a user has never opened a credit card, has no loan history, and has no repayments, they are considered white-account users, meaning they have no credit history. Traditional scoring models lack core assessment criteria for white-account users, so proxy variables are constructed using alternative data to supplement their credit profile.

[0074] Proxy variables primarily cover two areas: First, payment records for utility bills such as water, electricity, and gas, including payment continuity, any interruptions, and fluctuations in bill amounts. For example, if a user has paid their utility bills on time for 12 consecutive months, this indicates good residential stability and payment ability, positively reflecting their willingness to fulfill credit responsibilities. Second, data on e-commerce platform performance, including order completion rates, return rates, frequency of negative reviews, and post-sales dispute records. For example, if a user completed 50 e-commerce transactions in the past six months, returned only two items, and received positive reviews, this indicates stable transaction behavior and a strong willingness to fulfill contracts, which can be considered a positive proxy indicator of credit.

[0075] The credit graph unit 202 is used to determine financial behavior capabilities based on proxy variables, construct edge weights based on proxy variables to form a multidimensional credit graph, obtain existing data, identify communities with similar credit behaviors based on the existing data, and locate similar people.

[0076] In this module, the credit graph unit 202 determines financial capacity based on proxy variables. To assess a user's financial capacity in a pre-credit environment, it first quantitatively models their sense of responsibility, willingness to fulfill obligations, and lifestyle stability based on these proxy variables. These proxy variables include public payment records (such as the consistency and timeliness of utility bills), e-commerce fulfillment performance (such as order completion rates, return rates, and review content), telecommunications payment habits, and device location stability. By assigning weights and thresholds to each variable, a preliminary inference can be made as to whether a user possesses the financial capacity to fulfill obligations and make consistent payments on time. For example, a user who has made 12 consecutive months of payments, has an e-commerce return rate below 5%, and maintains a high and stable phone activity can be considered to have strong financial capacity.

[0077] Based on the aforementioned proxy variables, a multidimensional feature vector is constructed. A similarity metric (cosine similarity) is used to calculate the behavioral similarity between any two users. Users are represented as nodes in a graph structure, connected by edges, with edge weights reflecting the degree of behavioral similarity, thus forming a multidimensional credit graph. Once the graph is constructed, a graph embedding algorithm is used to map the nodes into a low-dimensional space. K-means clustering is then used to identify groups of users with similar characteristics, i.e., communities with similar credit behaviors. For example, if a majority of users in a community share the common characteristics of stable bill payments, good e-commerce performance, but no credit history, it can be inferred that the community's overall risk is low.

[0078] The contribution scoring unit 203 is used to obtain the contribution of existing white-household users, where the contribution includes continuous payment, e-commerce praise and unstable location, and determine the contribution score based on the contribution of the white-household users.

[0079] In this module, the contribution scoring unit 203 obtains the contribution of existing white-household users. To assess the matching quality and behavioral credibility of white-household users in the credit graph, the contribution score is calculated by calculating their performance on key proxy variable dimensions. Contribution measures the degree of consistency between white-household users' behavior and the characteristics of the community center to which they belong. It mainly includes three aspects: First, payment continuity, that is, whether the user has maintained regular payment of bills such as water, electricity, gas, and phone bills. If the user has made uninterrupted payments with little volatility for the past 12 months, their payment behavior is considered stable and has high positive contribution. Second, e-commerce positive reviews, which refers to the user's order completion rate, positive review rate, and low dispute rate on e-commerce platforms. A positive review rate above 90% indicates strong fulfillment capabilities and stable consumption behavior, which constitutes a clear positive credit signal. Third, location instability, which reflects whether the user's mobile device's geographical location frequently crosses regions or changes in a short period of time. Frequent changes of residence and unusual whereabouts are considered negative contribution factors, which may indicate unstable life or secretive behavior.

[0080] When constructing the contribution score, the aforementioned proxy variables are standardized and feature weights are assigned. A matching score is calculated based on their fit within the target community's central feature vector. For example, if a user's payment continuity and positive review rate are both highly similar to those of the target group, but their positional volatility is slightly higher, a contribution score of 0.76 might be output, indicating that their overall behavior is reliable but with slight deviations. Ultimately, this contribution score is used not only to determine a user's trustworthiness within the current community but also as a key indicator for their credit score weighting.

[0081] like Figure 7 As shown, as a preferred embodiment of the present invention, the risk identification module 300 includes: The real-time behavior unit 301 is used to obtain real-time behavior data streams, which include financial, e-commerce and public service data, and obtain platform data through the authorized API interface docking platform.

[0082] In this module, the real-time behavior unit 301 obtains real-time behavior data streams. To achieve dynamic identification and continuous updating of the user's credit status, a behavior evolution sequence is constructed by obtaining real-time behavior data streams. This data stream mainly includes three types of information: finance, e-commerce, and public services. The financial data section covers account balance changes, repayment behavior, credit usage, and overdue warnings. A typical example is that a user's failure to automatically repay a bank's loan triggers a negative behavior node; e-commerce data includes order completion status, evaluation records, return frequency, and shopping cycle. For example, three consecutive after-sales refunds will be considered a potential performance deviation signal; public service data refers to the timeliness and amount fluctuations of water, electricity, and gas bills, as well as whether there are sudden changes in urban commuting trajectories. For example, if a user stops paying their phone bills for two consecutive months or the device location signal frequently crosses provinces, it may indicate risk.

[0083] All data acquisition relies on standardized APIs provided by various platforms after user authorization. By establishing interfaces with banks, e-commerce platforms, and public utility service platforms, credit-related behavioral data is automatically pulled with user authorization and integrated into the user's credit history by timestamp. This process enables high-frequency collection and real-time synchronization of key credit behaviors, ensuring that the scoring model is dynamically updated based on the latest and most realistic behavioral status.

[0084] The behavior trajectory diagram unit 302 is used to construct a multi-dimensional credit behavior trajectory diagram based on the real-time behavior data stream and platform data. The trajectory diagram dimensions represent key credit factors and identify mutation nodes in the trajectory diagram.

[0085] In this module, the behavior trajectory diagram unit 302 constructs a multi-dimensional credit behavior trajectory diagram based on the real-time behavior data stream and platform data. Based on the real-time behavior data stream obtained from the financial, e-commerce and public service platforms, a multi-dimensional credit behavior trajectory diagram with users as the core is constructed. The trajectory diagram uses time as the horizontal axis and multiple key credit factors as the vertical dimensions. Each factor dimension corresponds to a behavior sequence curve updated by time, reflecting the changing trend of the user's credit behavior in different dimensions. Commonly used dimensions include repayment records, e-commerce performance, payment behavior, and position stability. For example, the continuity dimension of a user's water and electricity payment has remained stable in the past six months, but there have been two consecutive overdue payments in the current month, which will appear as an obvious mutation in the trajectory of this dimension.

[0086] To identify mutation nodes in the trajectory graph, statistical anomaly detection methods such as Z-score, sliding window mean deviation, or local outlier factor (LOF) are used to dynamically monitor each behavior curve. When the behavior value at a certain point in time deviates significantly from its historical average and crosses a set threshold, it is marked as a mutation node. For example, if a user's e-commerce return rate averaged 2% over the past 12 months but suddenly increased to 12% this month, this node would be identified as a mutation and used as an early signal of potential fulfillment risk for score adjustment.

[0087] The risk shift unit 303 is used to determine whether the behavior change constitutes a signal of credit risk shift based on the mutation node. If it is determined that there is a credit risk shift, the score is adjusted downward by calculating the correction factor and the node record is retained for trend monitoring.

[0088] In this module, the risk shift unit 303 determines whether the behavior change constitutes a signal of credit risk shift based on the mutation node. After the mutation node is identified in the multi-dimensional credit behavior trajectory diagram, it is necessary to further determine whether the behavior change constitutes a substantial credit risk shift signal. The judgment process comprehensively considers factors such as the mutation amplitude, duration, degree of deviation from the historical baseline, and credit sensitivity of the dimension to which it belongs. If the mutation occurs in a high-weight dimension, such as repayment behavior or e-commerce fulfillment, and the amplitude of the change exceeds the statistical threshold, it means that the behavior has major abnormal characteristics. Taking the user's repayment behavior in the past 12 months as an example, if the repayment has always been on time but has been delayed twice in a row this month, and the number of delay days exceeds 10 days, this behavior node will be marked as abnormal, and its deviation amplitude relative to the historical baseline will be further calculated. Combined with the importance weights of the behavioral dimensions, a correction factor is constructed.

[0089] like Figure 8 As shown, as a preferred embodiment of the present invention, the score interpretation module 400 includes: The score interpretation unit 401 is used to obtain the revised final score, quantify the positive or negative impact of each behavioral feature on the final score based on the final score, and visualize the quantification results.

[0090] In this module, the score interpretation unit 401 obtains the revised final score. After identifying mutation nodes and calculating the correction factor, the original score is superimposed with the correction factor to obtain the revised final score. This final score not only reflects the current user's overall credit status but also includes dynamic adjustments to short-term risk bias. To enhance the transparency and interpretability of the scoring results, the influence of each behavioral characteristic that contributes to the score is further quantified. An algorithm such as the SHAP (Shapley Additive Explanations) algorithm is used to perform inverse analysis of the scoring model and calculate the marginal contribution of each characteristic to the model's prediction results. Each behavioral factor is assigned a positive or negative scoring weight, indicating its contribution to or deduction from the final score. For example, a user's final score of 82 points might include a +10 contribution from utility bill continuity, a -6 contribution from career volatility, and a -4 contribution from e-commerce return rate. The system stores these quantitative results as scores, forming a score interpretation structure.

[0091] To enhance user understanding of the scoring structure, all feature contributions are visualized, often using bar charts, score composition diagrams, or stratified score distribution charts. The charts indicate the direction and strength of each factor's impact on the score. For example, a chart might show payment behavior and e-commerce fulfillment in the positive zone, while unstable device location is in the negative zone. This helps users quickly understand the primary factors contributing to their scores and identify potential behavioral optimization areas.

[0092] The material information updating unit 402 is used to obtain scoring components, which include credit enhancement factors, credit risk factors, and neutral features, and obtain updated evidence material information based on the scoring components. The evidence material information is used for scoring appeals.

[0093] In this module, the material information update unit 402 obtains the scoring components. While outputting the final score, it can further extract the scoring components and classify all the behavioral characteristics involved in the calculation of the score into three types: credit enhancement factors, credit risk factors, and neutral characteristics. Credit enhancement factors refer to behavioral characteristics that have a significant positive impact on the score, such as continuous payment of water, electricity, and gas bills, high e-commerce praise rate, and stable employment record. Credit risk factors are characteristics that have a negative impact on the score, such as frequent credit card overdue returns and frequent changes in location. Neutral characteristics are variables that have a weak impact on the score or do not fluctuate significantly, such as occasional small refunds or small payment delays. The scoring results are displayed in a structured manner, with the impact direction and impact weight of each factor attached, which is used for users to understand and locate complaints later.

[0094] When users have objections to the rating, they can focus on specific risk factors or key deduction items based on the above classification results and submit corresponding updated evidence materials. These materials may include payment receipts for phone bills that were mistakenly marked as arrears, records of e-commerce platform dispute cancellations, proof of position or address stability, etc. The system uses OCR and semantic analysis to identify the content of the materials and match the corresponding scoring elements, triggering a local re-scoring process after the model is verified. For example, if a user's rating is deducted due to overdue phone bills, and they provide valid payment screenshots and operator bills, the model comparison finds that it is indeed a mislabeling, and will automatically adjust the scoring factor value and update the scoring results.

[0095] The score correction unit 403 is configured to obtain a new score recommendation value based on the evidence material information. If the recommendation value is significantly different from the original score and the evidence material information is credible, the score is corrected and the score correction path is recorded.

[0096] In this module, the score correction unit 403 obtains a new score recommendation value based on the evidence material information. When the user submits the evidence material information related to the score result, the system will verify and update the corresponding score factor based on the material, and then re-run the score model to generate a new score recommendation value. The material verification process first extracts the voucher content through OCR text recognition, and then confirms its association with the specific score factor through semantic matching and field comparison algorithms. For example, a user is deducted points due to overdue phone bills. The uploaded phone bill payment record is identified and shown as automatic payment completed, and the payment time is exactly the same as the bill period. The model considers the material valid and accordingly corrects the value of the score factor from overdue to on-time payment. The score model recalculates the result after updating the factor. If the generated new score recommendation value is significantly different from the original score, such as exceeding the set threshold of 5 points, and the evidence material scores well in the credibility assessment, then the score recommendation will be adopted.

[0097] After the correction is completed, the scoring process will automatically record the complete path of the score adjustment, including the triggering factor, original value, modified value, material verification status, and the difference between the before and after scores, and generate a score correction log. For example, a user's initial score was 76 points, of which 6 points were mistakenly deducted due to an e-commerce dispute record due to a delayed system update. After the user submitted a revocation certificate and the model confirmed its validity, the factor was revised from dispute existing to dispute cleared, and the new score was updated to 81 points. The system archived this change as a score correction event in the user's credit history.

[0098] In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following steps are performed: Obtaining credit information data, where the credit information is obtained from multiple heterogeneous sources, including bank, consumer finance, and operator data; Based on the proxy variables, edge weights are constructed to form a multi-dimensional credit map. Based on the existing data, communities with similar credit behaviors are identified to locate similar groups of people. The contribution scores are determined based on the contribution of white household users. Build a multi-dimensional credit behavior trajectory diagram based on real-time behavioral data streams and platform data, identify mutation nodes in the trajectory diagram, and determine whether the behavioral change constitutes a signal of credit risk deviation based on the mutation nodes; The positive or negative impact of each behavioral characteristic on the final score is quantified based on the final score, updated evidence material information is obtained based on the scoring factors, and a new score recommendation value is obtained based on the evidence material information.

[0099] In one embodiment, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the processor performs the following steps: Obtaining credit information data, where the credit information is obtained from multiple heterogeneous sources, including bank, consumer finance, and operator data; Based on the proxy variables, edge weights are constructed to form a multi-dimensional credit map. Based on the existing data, communities with similar credit behaviors are identified to locate similar groups of people. The contribution scores are determined based on the contribution of white household users. Build a multi-dimensional credit behavior trajectory diagram based on real-time behavioral data streams and platform data, identify mutation nodes in the trajectory diagram, and determine whether the behavioral change constitutes a signal of credit risk deviation based on the mutation nodes; The positive or negative impact of each behavioral characteristic on the final score is quantified based on the final score, updated evidence material information is obtained based on the scoring factors, and a new score recommendation value is obtained based on the evidence material information.

[0100] It should be understood that, although the various steps in the flow chart of each embodiment of the present invention are shown in sequence according to the indication of the arrows, these steps are not necessarily performed in sequence according to the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in order, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0101] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When executed, the program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media used in the various embodiments provided herein may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM).

[0102] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0103] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

[0104] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A credit information data management and analysis method, characterized in that: The method comprises: Obtaining credit information data, where the credit information is obtained from multiple heterogeneous sources, including bank, consumer finance, and operator data; Based on the proxy variables, edge weights are constructed to form a multi-dimensional credit map. Based on the existing data, communities with similar credit behaviors are identified to locate similar groups of people. The contribution scores are determined based on the contribution of white household users. Build a multi-dimensional credit behavior trajectory diagram based on real-time behavioral data streams and platform data, identify mutation nodes in the trajectory diagram, and determine whether the behavioral change constitutes a signal of credit risk deviation based on the mutation nodes; The positive or negative impact of each behavioral characteristic on the final score is quantified based on the final score, updated evidence material information is obtained based on the scoring factors, and a new score recommendation value is obtained based on the evidence material information.

2. The credit information data management and analysis method according to claim 1, characterized in that: The steps of constructing edge weights based on proxy variables to form a multidimensional credit graph, identifying communities with similar credit behaviors based on existing data, locating similar groups of people, and determining contribution scores based on the contribution of white household users specifically include: Determine white household users based on credit data, where white household users have no credit history, and obtain proxy variables based on white household users, including water, electricity, and gas payment records and fulfillment rates and return rates on e-commerce platforms; Determine financial behavior capabilities based on proxy variables, construct edge weights based on proxy variables to form a multidimensional credit map, obtain existing data, identify communities with similar credit behaviors based on existing data, and locate similar groups of people; The contribution of existing white-household users is obtained, where the contribution includes continuous payment, e-commerce positive reviews, and unstable location, and a contribution score is determined based on the contribution of the white-household users.

3. The credit information data management and analysis method according to claim 1, characterized in that: The steps of constructing a multi-dimensional credit behavior trajectory diagram based on the real-time behavior data stream and platform data, identifying mutation nodes in the trajectory diagram, and determining whether the behavior change constitutes a signal of credit risk deviation based on the mutation nodes specifically include: Obtain real-time behavioral data streams, including financial, e-commerce, and public service data, and connect to the platform through authorized API interfaces to obtain platform data; Construct a multi-dimensional credit behavior trajectory map based on real-time behavior data streams and platform data. The dimensions of the trajectory map represent key credit factors, and identify mutation nodes in the trajectory map. Based on the mutation node, it is determined whether the behavioral change constitutes a signal of credit risk deviation. If it is determined that there is a credit risk deviation, the score is lowered by calculating the correction factor, and the node record is retained for trend monitoring.

4. The credit information data management and analysis method according to claim 1, characterized in that: The steps of quantifying the positive or negative impact of each behavioral feature on the final score based on the final score, obtaining updated evidence material information based on the scoring factors, and obtaining a new score recommendation value based on the evidence material information specifically include: Obtain the revised final score, quantify the positive or negative impact of each behavioral characteristic on the final score based on the final score, and visualize the quantified results; Obtaining scoring components, including credit enhancement factors, credit risk factors, and neutral characteristics, and obtaining updated evidentiary information based on the scoring components, which is used for scoring appeals; A new score recommendation value is obtained based on the evidence material information. If the recommendation value is significantly different from the original score and the evidence material information is credible, the score is revised and the score revision path is recorded.

5. The credit information data management and analysis method according to claim 3, characterized in that: The platforms include bank credit reporting, payment platforms, e-commerce platforms and operators.

6. A credit information data management and analysis system, characterized in that: The system comprises: A credit information module, which obtains credit data information from multiple heterogeneous sources, including bank, consumer finance, and operator data; The contribution scoring module constructs edge weights based on proxy variables to form a multi-dimensional credit map. Based on existing data, it identifies communities with similar credit behaviors, locates similar groups of people, and determines the contribution score based on the contribution of white household users. The risk identification module constructs a multi-dimensional credit behavior trajectory diagram based on real-time behavioral data streams and platform data, identifies mutation nodes in the trajectory diagram, and determines whether the behavioral change constitutes a signal of credit risk deviation based on the mutation nodes; The scoring interpretation module quantifies the positive or negative impact of each behavioral feature on the final score based on the final score, obtains updated evidence material information based on the scoring factors, and obtains a new score recommendation value based on the evidence material information.

7. The credit information data management and analysis system according to claim 6, characterized in that: The contribution scoring module includes: An agent variable unit determines white-account users based on credit data. The white-account users are users with no credit history, and obtains agent variables based on the white-account users. The agent variables include water, electricity, and gas payment records and fulfillment rates and return rates on e-commerce platforms. The credit graph unit determines financial behavior capabilities based on proxy variables, constructs edge weights based on proxy variables to form a multi-dimensional credit graph, obtains existing data, identifies communities with similar credit behaviors based on the existing data, and locates similar groups of people; The contribution scoring unit obtains the contribution of existing white-household users, where the contribution includes continuous payment, e-commerce praise and unstable location, and determines the contribution score based on the contribution of the white-household users.

8. The credit information data management and analysis system according to claim 7, characterized in that: The risk identification module includes: The real-time behavior unit obtains real-time behavior data streams, including financial, e-commerce, and public service data, and obtains platform data through the authorized API interface connection platform; The behavior trajectory graph unit constructs a multi-dimensional credit behavior trajectory graph based on real-time behavior data streams and platform data. The trajectory graph dimensions represent key credit factors and identify mutation nodes in the trajectory graph. The risk shift unit determines whether the behavioral change constitutes a signal of credit risk shift based on the mutation node. If it is determined that there is a credit risk shift, the score is lowered by calculating the correction factor and the node record is retained for trend monitoring.

9. The credit information data management and analysis system according to claim 8, characterized in that: The score explanation module includes: The score interpretation unit obtains the revised final score, quantifies the positive or negative impact of each behavioral characteristic on the final score based on the final score, and visualizes the quantified results; A material information updating unit, which obtains scoring components, including credit enhancement factors, credit risk factors, and neutral characteristics, and obtains updated evidence material information based on the scoring components. The evidence material information is used for scoring appeals; The scoring correction unit obtains a new scoring recommendation value based on the evidence material information. If the recommendation value is significantly different from the original score and the evidence material information is credible, the score is corrected and the scoring correction path is recorded.

10. A credit information data management and analysis system according to claim 9, characterized in that: The platforms include bank credit reporting, payment platforms, e-commerce platforms and operators.

Images