A user big data processing method based on supply chain finance

Abstract

The application discloses a user big data processing method based on supply chain finance, which comprises the following steps: collecting user interaction data, extracting short-term time sequence and long-term behavior characteristics to construct an interaction data network, initializing weights of risk-related characteristics and training and adjusting; then, calculating a user preference decay coefficient according to a transaction time interval, segmenting behavior characteristics, identifying behavior and preference changes in each stage, evaluating the influence of risks and adjusting the weights; subsequently, constructing an interaction data operation chain based on the data network and adjusting the behavior change rate of the user in different stages; finally, constructing a risk transmission rule and grouping according to historical risk characteristics and determining a key behavior change rate; through multi-dimensional feature analysis and behavior change rate calculation, the user dynamics are accurately described, potential risk correlations are mined, an evaluation model is constructed in combination with historical data, risk signals can be captured in time, the accuracy and timeliness of risk identification are improved, and the platform risk management is facilitated.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a user big data processing method based on supply chain finance. Background Art

[0002] In the field of supply chain finance, credit data traditionally relies primarily on static financial reports, leading to prominent information asymmetry, limited credit granting methods, and slow data updates. This, in turn, leaves small and medium-sized enterprises facing financing difficulties and significant financial pressure. As market competition shifts from single-customer competition to supply chain competition, the problem of asynchronous capital flows between enterprises has become increasingly prominent.

[0003] At the same time, big data technology is rapidly developing. Its characteristics, such as massive data volumes, rapid circulation, diverse data types, and low value density, offer new solutions to the challenges of supply chain finance. Big data technology can integrate multi-source data to create precise profiles of small and medium-sized enterprises (SMEs), monitor their dynamic operating data in real time, and cross-validate detailed transaction records such as orders and inventory to gain insights into the true state of a company's operations, improve the quality of credit reporting services, and mitigate risks. Furthermore, big data technology can enable real-time early warning and quantitative credit granting, eliminating reliance on rigid guarantees from core enterprises, alleviating financing difficulties for SMEs and promoting the digital transformation of supply chain finance.

[0004] With the booming development of supply chain finance, accurately assessing user risk is crucial for ensuring stable business operations and preventing financial risks. However, traditional user risk assessment methods have significant shortcomings. Supply chain finance platforms currently accumulate vast amounts of user interaction data, but traditional methods often focus on only a subset of static data, failing to fully capture the dynamics of user behavior. Furthermore, they fail to fully consider the complex correlations between different characteristics and their differential impact on risk. Summary of the Invention

[0005] This application provides a user big data processing method based on supply chain finance to more accurately assess user risks. It is necessary to collect all user interaction data on the supply chain finance platform, extract short-term time series and long-term behavioral characteristics to build an interactive data network, and then learn to adjust risk weights, calculate preference attenuation coefficients, evaluate the risk impact of each stage, and build an interactive data operation chain and risk transmission rules to achieve accurate control of user risks.

[0006] This application provides a user big data processing method based on supply chain finance, including:

[0007] S1, collects all interaction data of users on the supply chain finance platform, extracts users' short-term time series features and long-term behavior features to form an interaction data network;

[0008] S2, selects features related to user risk from the interaction data relationship network to initialize risk weights, trains models based on historical features, and learns and adjusts the weights of different features on user risk;

[0009] S3, based on the interactive data network between transaction time and current time interval, calculates the decay coefficient of user preference over time and performs behavioral feature segmentation;

[0010] S4, identifying the user's behavioral characteristics and preference changes at each stage, assessing the risk impact of the user's behavioral characteristics at different stages and adjusting the risk weight;

[0011] S5, builds an interactive data operation chain based on the interactive data network to adjust the user's behavior change rate at different stages;

[0012] S6, construct risk transmission rules and groupings based on historical risk characteristics, and determine the change rate of key behaviors before risk data.

[0013] Preferably, the interactive data network includes: collecting the interactive data of all users, treating each user as a node, and also treating the extracted short-term time series features and long-term behavioral features as nodes, constructing weighted edges based on the correlation between the features and visually displaying them.

[0014] Preferably, the short-term time series features and long-term behavior features specifically include that the short-term time series features are the interaction data of the user within a time window, and the long-term behavior features are the interaction data of the user within n historical time windows.

[0015] Preferably, the feature segmentation specifically includes: obtaining all transaction records of the user from the database of the supply chain finance platform, including the timestamp of each transaction, and calculating the interval between each transaction and the current time based on the current time; for each transaction, And the selected decay function, calculate the decay coefficient corresponding to the transaction , where β is the attenuation factor; according to the size of the attenuation coefficient, the segmentation threshold is set.

[0016] Preferably, the adjustment of risk weights specifically includes: extracting key behavioral indicators from the behavioral characteristic data of each stage, clustering the transaction data of each stage using K-means clustering, using behavioral characteristics as input variables for clustering, classifying similar transaction behaviors into one category, and finding the correlation between transaction behaviors; comparing behavioral characteristics of different stages and analyzing changes in user preferences; collecting historical risk data and associating the behavioral characteristics of each stage with historical risk data, analyzing the correlation between different behavioral characteristics and risk events; calculating the risk probability of each behavioral characteristic based on the correlation between the behavioral characteristics and risk events, and adjusting the initialized risk weight based on the risk probability.

[0017] Preferably, the initializing risk weight includes: screening out features closely related to user risk from the conversation interaction data network, randomly generating an initial weight value for each selected feature, and setting the initial weight for the feature based on business experience.

[0018] Preferably, the adjusting the initialization risk weight includes: adjusting the calculation formula of the initialization risk weight to: , is the initial risk weight of behavioral characteristic B, α is the adjustment coefficient, is the risk probability of each behavioral feature.

[0019] Preferably, the adjustment of the user's behavior change rate at different stages is specifically as follows: sorting the collected interaction data in the interactive data network in ascending order according to the operation time to form a preliminary operation sequence; integrating the operation sequence after sorting and associating the operation objects into a complete interactive data operation chain; analyzing the behavior change rate of the same type of interactive data operation chain at different time stages; and formulating an adjustment strategy according to the size and positive and negative of the behavior change rate Rc.

[0020] Preferably, the change rate Rc includes: ,in, It is the behavioral characteristic of the previous stage. It is a characteristic of the latter stage. is the risk weight adjusted for the behavioral characteristics of the previous stage, It is the risk weight adjusted for the behavioral characteristics of the latter stage.

[0021] Preferably, the determination of the key behavior change rate before risk data specifically includes: collecting risk event records and related transaction record data of various types of interactive data operation chains from the historical database of the supply chain finance platform; extracting characteristic indicators of each risk type corresponding to each type of interactive data operation chain from the collected data; analyzing the correlation between each type of interactive data operation chain and the risk type, finding out the combination with risk transmission relationship, and placing the interactive data operation chains and risk types with risk transmission relationship in the same grouping; screening out the interactive data operation chain type and risk type combination that exceeds the threshold, and for each grouping, collecting historical interactive data features before the risk occurs in the interactive data operation chain process, determining the behavior change rate before each historical interactive data feature, and establishing a risk assessment model based on the behavior change rate before the historical interactive data feature.

[0022] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0023] By collecting user interaction data and building an interactive data network, we extract multi-dimensional features and calculate behavioral change rates to accurately characterize user behavior and preference dynamics. We construct groupings based on feature correlations and risk transmission patterns, enabling in-depth exploration of potential risk correlations. By adjusting risk weights based on historical risk data and constructing a risk assessment model, we can promptly capture risk signals from changing user behavior. By comprehensively considering the temporal dimension of user behavior and its feature correlations, we improve the accuracy and timeliness of risk identification, providing strong support for risk management within supply chain finance platforms.

[0024] By setting a behavioral change rate range for the interactive data operation chain and simulating its changes, the simulated change rate is precisely matched to the risk characteristics, thereby calculating the risk probability and assessing risk resilience. This solution can comprehensively and dynamically reflect the risk status of the interactive data operation chain under different states. This solution can proactively identify potentially high-risk operation chains and behavioral change states, deeply tap into the value of user data, and accurately capture behavioral changes in user interactive data operation chains in supply chain finance scenarios. It can provide timely insights into potential risks and provide early warnings of high-risk operation chains and behavioral states, helping financial institutions effectively mitigate credit and market risks. It also helps optimize financial service strategies, provide personalized financial solutions for users with different risk profiles, improve the quality and efficiency of financial services, enhance the stability and sustainability of supply chain finance businesses, and promote the healthy development of the supply chain finance ecosystem. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 The figure is a flow chart of a method for processing user big data based on supply chain finance according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] To facilitate understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to enable a more thorough and comprehensive understanding of the disclosed content of the present invention.

[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains; the terms used herein in the specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more of the associated listed items.

[0028] Example 1: Figure 1 This is a flow chart of a user big data processing method based on supply chain finance according to an embodiment of the present invention.

[0029] like Figure 1 As shown, a user big data processing method based on supply chain finance includes the following steps:

[0030] S1, collects all interaction data of users on the supply chain finance platform, extracts users' short-term time series features and long-term behavioral features to form an interaction data network.

[0031] Short-term time series features are user interaction data within a single time window (a week or a month). Long-term behavioral features are user interaction data within n historical time windows (the past year or two). Interaction data includes, but is not limited to, user login time, pages viewed, transaction history (borrowing, repayment, and payment), query history (account balance and transaction details inquiries), click behavior, and dwell time.

[0032] Specifically, we collect interaction data from all users on the supply chain finance platform's log system, database, and user behavior tracking tools. For short-term time series features, we extract login and transaction frequency features (daily logins and weekly transactions), amount features (total transaction amount in the past week, maximum single transaction amount, etc.), time features (last login time, last transaction time, etc.), and behavior sequence features (such as the user's behavior sequence in the past week (login-browsing-transaction-logout, etc.)). For long-term behavior features, we extract cumulative features (annual total transaction amount, average annual number of transactions, etc.), trend features (transaction amount growth trend, transaction number change trend, etc.), and stability features (transaction time stability (whether transactions are fixed on certain days), transaction amount volatility, etc.).

[0033] Each user is considered a node, and the extracted short-term time series features and long-term behavioral features are also considered nodes. Weighted edges are constructed based on the correlation between features (such as the frequency of co-occurrence, time interval, causal relationship, etc.). For example, if a user's short-term transaction amount is positively correlated with the growth trend of their long-term transaction amount, a positively weighted edge is constructed between them. The correlation weight between feature nodes is calculated using cosine similarity (the formula is: , where A and B are vector representations of two feature nodes, and n is the dimension of the vector. Visualization tools are used to visualize the conversation interaction data network, making it easier to intuitively understand the relationship between user features.

[0034] S2, selects features related to user risk from the interactive data relationship network to initialize risk weights, trains models based on historical features, and learns and adjusts the weights of different features affecting user risk.

[0035] Relevant features include, but are not limited to, transaction amount, transaction frequency, counterparty credit rating, overdue records, query frequency (especially risk-related queries, such as credit report inquiries), abnormal login behavior, etc.

[0036] Specifically, we screen features closely related to user risk from the conversation interaction data network, randomly generate an initial weight for each selected feature, and set the initial weights for each feature based on business experience. For example, if the transaction amount is considered to have a greater impact on risk, its initial weight can be set relatively high; while for less important features, the initial weights can be set lower.

[0037] S3, based on the interactive data network between transaction time and current time interval, calculates the attenuation coefficient of user preference changes over time and performs behavioral feature segmentation.

[0038] Specifically, all transaction records of users are obtained from the database of the supply chain finance platform, including the timestamp of each transaction. Based on the current time, the interval between each transaction and the current time is calculated. For example, if the current time is , the time of a transaction is , then the time interval of the transaction We consider each transaction as a node and the transaction time interval as a node attribute to construct a transaction time interval interaction data network. This can be represented as a graph structure, where nodes represent transactions and edges represent connections between transactions (e.g., transactions of the same user may be connected by edges).

[0039] For each transaction, according to its time interval And the selected decay function, calculate the decay coefficient corresponding to the transaction , where β is the decay factor, controlling the rate of decay. The value of β can be determined through historical data analysis and business experience. For example, analyze user behavior changes over time and try different β values ​​to select the one that best fits the model.

[0040] According to the size of the attenuation coefficient, set the segmentation threshold. For example, you can set two thresholds and ( > ), the attenuation coefficient is divided into three intervals, ( ,1]、( , ] and [0, ]. Based on the decay coefficient of each transaction and the set threshold, the transaction is divided into different stages. The decay coefficient is ( ,1] are classified as the recent stage, indicating that these transactions have a greater impact on the user's current preferences; the attenuation coefficient is ( , ] is divided into the mid-term stage, and the impact is relatively small; the attenuation coefficient is [0, ] is divided into the long-term stage with the least impact. For each stage, the behavioral characteristics corresponding to the transactions in that stage are aggregated. For example, the total transaction amount, average transaction amount, number of transactions, etc. in that stage are calculated. The total transaction amount of the behavioral characteristic aggregation formula is: for k transactions in a certain stage, the total transaction amount ,in is the amount of the jth transaction. The average transaction amount is: , number of transactions N=k.

[0041] S4, identify the user's behavioral characteristics and preference changes at each stage, evaluate the risk impact of the user's behavioral characteristics at different stages and adjust the risk weight.

[0042] Among them, the key behavioral characteristics are transaction frequency, transaction amount, transaction object type (supplier, buyer, etc.), transaction time distribution, etc.

[0043] Specifically, we extract key behavioral indicators from the behavioral characteristics of each phase and cluster the transaction data for each phase using K-means clustering. Using behavioral characteristics as clustering input variables, we group similar transactions together. For example, in the recent phase, we might cluster different behavioral characteristics, such as "high-frequency small-value transactions" and "low-frequency large-value transactions."

[0044] Association rule mining algorithms are used to identify relationships between transaction behaviors. For example, if a user purchases a specific product and then subsequently makes another type of transaction, this may reflect a user's preference or business need.

[0045] Compare behavioral characteristics across different phases to analyze changes in user preferences. For example, if a user has recently begun frequently transacting with new suppliers, whereas previously they primarily worked with fixed suppliers, this may indicate a shift in their purchasing preferences.

[0046] Collect historical risk data (including records of risk events (such as overdue payments and defaults) experienced by users at different stages) and correlate the behavioral characteristics of each stage with the historical risk data to analyze the correlation between different behavioral characteristics and risk events. For example, calculate the proportion of users with the "high-frequency small-amount transaction" characteristic among those who experienced risk events and those who did not.

[0047] Based on the correlation between behavioral characteristics and risk events, the risk probability of each behavioral characteristic is calculated, and the initial risk weight is adjusted based on the risk probability. For example, the risk probability P(R|B) of each behavioral characteristic is calculated, that is, the probability of the user experiencing risk event R under the condition that behavior pattern B occurs. The calculation formula is:

[0048]

[0049] Among them, n(B∩R) is the number of users who have both behavioral feature B and risk event R, and n(B) is the number of users who have behavioral feature B.

[0050] Using the linear weighting method, the risk probability is combined with the initial weight, and the risk weights of all behavioral characteristics are normalized so that their sum is 1. For example, for behavioral characteristic B, the adjusted risk weight is It can be expressed as:

[0051]

[0052] is the initial risk weight of behavioral characteristic B, and α is the adjustment coefficient, which can be adjusted according to actual conditions.

[0053] S5, builds an interactive data operation chain based on the interactive data network to adjust the user's behavior change rate at different stages.

[0054] Among them, the interactive data operation chain is a data sequence that records a series of operations performed by users on the supply chain finance platform in chronological order, which is used to reflect the user's behavioral trajectory on the platform.

[0055] Specifically, the collected interaction data in the interactive data network is sorted in ascending order by operation time to form a preliminary operation sequence (for example, user A's operation records include: Operation 1 (login, time 2024-01-01 09:00:00), Operation 2 (query transaction records, time 2024-01-01 09:05:00), and Operation 3 (submit order, time 2024-01-01 09:10:00). After sorting, the sequence becomes Operation 1 → Operation 2 → Operation 3). For data involving operation objects, the operation is associated with the corresponding operation object (for example, if the order submitted by Operation 3 is numbered 001, Operation 3 (submit order, order number 001) is clearly recorded in the interactive data operation chain). The operation sequence, after sorting and associating the operation objects, is integrated into a complete interactive data operation chain.

[0056] Analyze the behavioral change rates of similar interactive data operation chains at different time stages, such as the behavioral change rates of the favorite item interactive data operation chain and the behavioral change rates of the order item interactive data operation chain. Develop appropriate adjustment strategies based on the magnitude and sign of the behavioral change rate (Rc). For example, if Rc is positive and has a large absolute value, it indicates that user behavior has undergone significant positive changes during this period, potentially indicating business expansion. However, potential risks, such as increased financial pressure, require attention. In this case, the user's risk assessment threshold can be appropriately raised, while also strengthening monitoring of cash flow. If Rc is negative and has a large absolute value, it indicates that user behavior has undergone significant negative changes during this period, potentially indicating business contraction or problems. In this case, the user's risk assessment threshold should be lowered and investigation and analysis of the user's business status should be strengthened, such as contacting the user to understand the cause and checking transaction records for anomalies.

[0057] The behavior change rate calculation formula is:

[0058]

[0059] in, It is the behavioral characteristic of the previous stage. It is a characteristic of the latter stage. is the risk weight adjusted for the behavioral characteristics of the previous stage, It is the risk weight adjusted for the behavioral characteristics of the latter stage.

[0060] S6, construct risk transmission rules and groupings based on historical risk characteristics, and determine the change rate of key behaviors before risk data.

[0061] Specifically, risk event records and related transaction record data of various interactive data operation chains are collected from the historical database of the supply chain finance platform, including the time of risk occurrence, business links involved (such as procurement, production, sales, logistics, settlement, etc.), risk type (such as credit risk, market risk, operational risk, liquidity risk, etc.), amount of risk loss and other information.

[0062] The characteristic indicators of each risk type corresponding to each type of interactive data operation chain are extracted from the collected data.

[0063] Analyze the relationship between each type of interactive data operation chain and risk type, find out the combination with risk transmission relationship, and put the interactive data operation chains and risk types with risk transmission relationship in the same group.

[0064] Specifically, we collect and clean the interactive data operation chain records and risk event records, and integrate them to form a data set containing interactive data operation chain types and corresponding risk types. We create a matrix with rows representing interactive data operation chain types and columns representing risk types. The elements in the matrix represent the strength of the association between a specific interactive data operation chain type and a specific risk type. We calculate the number of occurrences of each combination through frequency statistics, and then use the conditional probability formula ( ) measures the likelihood of occurrence of a specific risk type under the interactive data operation chain, and uses the mutual information formula to evaluate the statistical dependence between the two to quantify the strength of the association.

[0065] Based on business needs and correlation strength, we set a correlation strength threshold, screen out combinations of interactive data operation chain types and risk types that exceed the threshold, analyze the causal or time series relationships between these combinations, and determine whether there is a risk transmission relationship. Based on the results of the risk transmission relationship analysis, combinations with risk transmission relationships are grouped together.

[0066] Among them, the mutual information formula is:

[0067]

[0068] Represents the type of interactive data operation chain of type i, which is a discrete random variable whose value set is All possible interactive data operation chain types in. For example, in supply chain finance, It may represent a chain of interactive data operations such as "order processing". Represents the jth risk type, which is also a discrete random variable with a value set of For example, It can be "credit risk". A specific value of , that is, a specific instance of the interactive data operation chain type. For example, a specific order processing process in the interactive data operation chain of "order processing". r is A specific value of , that is, a specific instance of a risk type. For example, a specific credit default event in "credit risk"; P(o,r) represents the joint probability of interaction data operation chain type o and risk type r, that is, the probability that interaction data operation chain type o and risk type r occur simultaneously. For example, the probability of credit default events occurring simultaneously in all order processing processes. P(o) represents the marginal probability of interaction data operation chain type o, that is, the probability that interaction data operation chain type o occurs. For example, the probability of all order processing processes occurring. P(r) represents the marginal probability of risk type r, that is, the probability that risk type r occurs. For example, the probability of all credit default events occurring.

[0069] For each grouping, historical interaction data features before the risk occurs in the interaction data operation chain process are collected, a behavior change rate before each historical interaction data feature is determined, and a risk assessment model is established according to the behavior change rate before the historical interaction data feature.

[0070] The historical interaction data includes transaction records, operation records, etc.

[0071] The technical solutions in the embodiments of the present application have at least the following technical effects or advantages:

[0072] By collecting user interaction data to construct an interaction data network, extracting multi-dimensional features and calculating behavior change rates, user behavior and preference dynamics are accurately described. Based on feature association and risk transmission rules, groupings are constructed, which can deeply mine potential risk associations. Combined with historical risk data to adjust risk weights, a risk assessment model is constructed, which can capture risk signals in user behavior changes in time, fully consider the time dimension of user behavior and feature association, improve the accuracy and timeliness of risk identification, and provide strong support for risk management of the supply chain finance platform.

[0073] Embodiment two: in embodiment one, for user big data processing in the supply chain finance scenario, although a risk analysis framework based on interaction data operation chain is initially constructed, in actual application, the framework considers the change rate of different risk features more single, and cannot fully combine the association between risk features and the complexity of business scenarios. In the face of different types of users, different business links, and the diversity of interaction data operation chain behavior change rates, it is difficult to accurately reflect the actual risk situation by only judging the risk probability according to the simple threshold, there is a high risk of misjudgment, and the risk assessment strategy cannot be dynamically adjusted according to different risk feature combinations. In order to more accurately evaluate the risk of user interaction data operation chain in supply chain finance and improve the accuracy and effectiveness of risk early warning, the existing method is optimized and improved.

[0074] In some embodiments, the key behavior change rate before the risk data is determined, and step S6 further includes:

[0075] S61, for each interaction data operation chain in the grouping, set different behavior change rate ranges and simulate the behavior change rate of the interaction data operation chain.

[0076] Specifically, we analyze and identify key risk-influencing characteristics within the interactive data operation chains within the group, such as transaction amount, transaction frequency, and the behavioral change rate of counterparty credit scores. For each key risk characteristic, we set a reasonable behavioral change rate range based on business experience and historical data fluctuations. For each interactive data operation chain within the group, we select all key risk characteristics and randomly select a behavioral change rate value within the set behavioral change rate range. Based on this selected behavioral change rate value, we simulate changes to the data for the corresponding characteristic within the interactive data operation chain.

[0077] S62, matching the simulated behavior change rate with the risk characteristics, and calculating the value of the behavior change rate of each interactive data operation chain data on each risk characteristic.

[0078] Specifically, analyze the correlation between the behavior change rate of each interactive data operation chain and each risk feature. Based on the analysis results, establish a mapping table between risk features and behavior change rates. The mapping table should include the name of the interactive data operation chain, behavior change rate indicators (such as request frequency behavior change rate, data volume behavior change rate, etc.) and the corresponding risk features. From the results of simulating the behavior change rate of the interactive data operation chain, obtain the behavior change rate data of each interactive data operation chain in different time intervals. According to the mapping table between risk features and behavior change rates, traverse the behavior change rate indicators and corresponding risk features of each interactive data operation chain. For each matching relationship, associate the behavior change rate data of the interactive data operation chain with the risk feature. Set a corresponding threshold for each risk feature, compare the behavior change rate of the interactive data operation chain with the threshold, and determine the value of the behavior change rate on the risk feature based on the comparison result.

[0079] S63, calculating the risk probability of each simulated interactive data operation chain under different behavior change rate states according to the value of the behavior change rate.

[0080] Specifically, from the results of the behavior change rate of the simulated interactive data operation chain, the behavior change rate values ​​of each simulated interactive data operation chain at different time intervals or states are obtained. The behavior change rate data of the simulated interactive data operation chain is input into the risk assessment model. Based on the input behavior change rate data, the model calculates the risk probability of each simulated interactive data operation chain under different behavior change rate states. The calculation formula for risk probability P is ,in, is the bias term, , ,⋯, is the weight coefficient, , ,⋯, is the rate of change of each behavior. Analyze the calculated risk probability to identify high-risk interactive data operation chains and behavior change rate states.

[0081] S64, evaluating the risk resistance of the interactive data operation chain under different behavior change rate states based on the risk probability, and determining the risk threshold interval to which the user belongs based on the risk probability corresponding to the user's current risk defense capability, and issuing a risk warning.

[0082] Among them, risk resistance refers to the ability of the interactive data operation chain to resist risks under different behavior change rates.

[0083] Specifically, we divide risk threshold intervals into categories based on business risk tolerance and historical risk data, collect behavioral change rate data of the user's current interactive data operation chain, and use the risk probability calculation formula to input the user's current behavioral change rate data into the model to calculate the user's current risk probability P. Based on the calculated risk probability P, we determine the risk threshold interval to which the user belongs, and issue a risk warning based on the judgment result.

[0084] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0085] By setting a behavioral change rate range for the interactive data operation chain and simulating its changes, the simulated change rate is precisely matched to the risk characteristics, thereby calculating the risk probability and assessing risk resilience. This solution can comprehensively and dynamically reflect the risk status of the interactive data operation chain under different states. This solution can proactively identify potentially high-risk operation chains and behavioral change states, deeply tap into the value of user data, and accurately capture behavioral changes in user interactive data operation chains in supply chain finance scenarios. It can provide timely insights into potential risks and provide early warnings of high-risk operation chains and behavioral states, helping financial institutions effectively mitigate credit and market risks. It also helps optimize financial service strategies, provide personalized financial solutions for users with different risk profiles, improve the quality and efficiency of financial services, enhance the stability and sustainability of supply chain finance businesses, and promote the healthy development of the supply chain finance ecosystem.

[0086] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Various modifications and variations are readily apparent to those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A user big data processing method based on supply chain finance, characterized in that: include: S1, collects all interaction data of users on the supply chain finance platform, extracts users' short-term time series features and long-term behavior features to form an interaction data network; S2, selects features related to user risk from the interaction data relationship network to initialize risk weights, trains models based on historical features, and learns and adjusts the weights of different features on user risk; S3, based on the interactive data network between transaction time and current time interval, calculates the decay coefficient of user preference over time and performs behavioral feature segmentation; S4, identifying the user's behavioral characteristics and preference changes at each stage, assessing the risk impact of the user's behavioral characteristics at different stages and adjusting the risk weight; S5, builds an interactive data operation chain based on the interactive data network to adjust the user's behavior change rate at different stages; S6, construct risk transmission rules and groupings based on historical risk characteristics, and determine the change rate of key behaviors before risk data; The S3 includes obtaining all transaction records of users from the database of the supply chain finance platform, including the timestamp of each transaction, and calculating the interval between each transaction and the current time based on the current time; for each transaction, according to its time interval And the selected decay function, calculate the decay coefficient corresponding to the transaction ,in is the attenuation factor; based on the size of the attenuation coefficient, the segmentation threshold is set; from the behavioral feature data of each stage, key behavioral indicators are extracted, and the transaction data of each stage is clustered using K-means clustering. The behavioral features are used as the input variables for clustering, similar transaction behaviors are grouped together, and the correlation between transaction behaviors is identified; Said S4 includes comparing the behavioral characteristics of different stages and analyzing the changes in user preferences; Collect historical risk data and correlate the behavioral characteristics of each stage with historical risk data, and analyze the correlation between different behavioral characteristics and risk events; Based on the correlation between behavioral characteristics and risk events, the risk probability of each behavioral characteristic is calculated, and the initial risk weight is adjusted according to the risk probability; In the interactive data network, the collected interactive data are sorted in ascending order according to the operation time to form a preliminary operation sequence; Integrate the operation sequence after sorting and associating the operation objects into a complete interactive data operation chain. Analyze the behavior change rate of the same type of interactive data operation chain at different time stages, and formulate adjustment strategies based on the size and positive and negative of the behavior change rate Rc. The formula for calculating Rc is: ,in, It is the behavioral characteristic of the previous stage. It is a characteristic of the latter stage. is the risk weight adjusted for the behavioral characteristics of the previous stage, is the risk weight adjusted for behavioral characteristics in the latter stage; The determination of the key behavior change rate before risk data specifically includes: collecting risk event records and related transaction record data of various types of interactive data operation chains from the historical database of the supply chain finance platform; extracting characteristic indicators of each risk type corresponding to each type of interactive data operation chain from the collected data; analyzing the correlation between each type of interactive data operation chain and the risk type, finding out the combination with risk transmission relationship, and placing the interactive data operation chains and risk types with risk transmission relationship in the same group; screening out the interactive data operation chain type and risk type combination that exceeds the threshold, for each group, collecting historical interactive data features before the risk occurs in the interactive data operation chain process, determining the behavioral change rate of each historical interactive data feature, and establishing a risk assessment model based on the behavioral change rate before the historical interactive data feature.

2. The user big data processing method based on supply chain finance according to claim 1 is characterized in that: The interactive data network includes: collecting the interactive data of all users, treating each user as a node, and also treating the extracted short-term time series features and long-term behavioral features as nodes, constructing weighted edges based on the correlation between the features, and visually displaying them.

3. The user big data processing method based on supply chain finance according to claim 2 is characterized in that: The short-term time series features and long-term behavior features specifically include: short-term time series features are the interaction data of the user within a time window; long-term behavior features are the interaction data of the user within n historical time windows.

4. The user big data processing method based on supply chain finance according to claim 1 is characterized in that: Initializing the risk weight includes: selecting features closely related to user risk from the conversation interaction data network, randomly generating an initial weight value for each selected feature, and setting the initial weight for the feature based on business experience.

5. The user big data processing method based on supply chain finance according to claim 1 is characterized in that: The adjusting the initialization risk weight includes: adjusting the calculation formula of the initialization risk weight to: , is the initial risk weight of behavioral characteristic B, α is the adjustment coefficient, is the risk probability of each behavioral feature.

Images