Payment account reconciliation early warning method and system based on multi-dimensional data

Abstract

The application discloses a payment reconciliation early warning method and system based on multidimensional data, and belongs to the technical field of reconciliation, which comprises the following steps: monitoring the transaction conditions of a payment system and a bank system, and collecting multidimensional data of the payment system and the bank reconciliation; processing the multidimensional data of the payment system and the bank reconciliation, and determining characteristic data of the payment system and the bank reconciliation; constructing a payment reconciliation anomaly detection model to analyze the characteristic data of the payment system and the bank reconciliation, automatically identifying abnormal conditions of the payment system and the bank system in the payment reconciliation, and automatically triggering an early warning alarm when abnormal conditions exist. The application solves the problem that the existing payment system and bank reconciliation cannot be monitored and abnormal conditions in the reconciliation cannot be found, and the reconciliation efficiency is low. The application can effectively monitor the reconciliation of the payment system and the bank, can timely find abnormal conditions of the payment system and the bank system in the payment reconciliation, can improve the reconciliation efficiency, and can fully guarantee the safety of funds.

Description

Technical Field

[0001] This invention relates to the field of reconciliation technology, specifically to a payment reconciliation early warning method and system based on multi-dimensional data. Background Technology

[0002] After a transaction occurs between the payment system and the banking system, reconciliation between the two systems is necessary. This reconciliation is a crucial step in ensuring financial accuracy, fund security, and risk control. By verifying account balances, transaction details, and other information, errors can be identified and corrected promptly, preventing financial problems caused by discrepancies in accounts. It also helps to detect unauthorized transactions, erroneous operations, or fraudulent activities, allowing for timely measures to prevent financial losses.

[0003] Existing technologies cannot effectively monitor the reconciliation between payment systems and banks, cannot promptly detect anomalies in payment reconciliation, reduce reconciliation efficiency, and cannot fully guarantee fund security. Summary of the Invention

[0004] The purpose of this invention is to provide a payment reconciliation early warning method and system based on multi-dimensional data, which can effectively monitor the reconciliation between the payment system and the bank, promptly detect abnormalities in the payment reconciliation between the payment system and the bank system, improve reconciliation efficiency, fully protect fund security, and solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Payment reconciliation early warning methods based on multi-dimensional data include:

[0007] Monitor transaction activity in payment and banking systems, and collect multi-dimensional reconciliation data from payment and banking systems;

[0008] The multi-dimensional data of payment system and bank reconciliation are processed, feature vectors are extracted and weighted and fused to determine the feature data of payment system and bank reconciliation.

[0009] A payment reconciliation anomaly detection model is constructed and the reconciliation feature data of the payment system and bank are analyzed to automatically identify anomalies in the payment reconciliation of the payment system and bank system, determine the payment reconciliation anomaly detection results, and automatically trigger early warning alarms when anomalies are found.

[0010] Preferably, the system automatically identifies anomalies in payment reconciliation between the payment system and the banking system and performs the following operations:

[0011] Deploy the optimal payment reconciliation anomaly detection model and place it in a real payment reconciliation anomaly detection environment;

[0012] The payment system and bank reconciliation feature data are input into the payment reconciliation anomaly detection model. The model analyzes the payment system and bank reconciliation feature data and automatically identifies anomalies in the payment system and bank system during payment reconciliation, thereby determining the payment reconciliation anomaly detection result.

[0013] When anomalies are found in the payment system and banking system during payment reconciliation, an early warning alarm is automatically triggered, the source of the anomaly is quickly traced, and administrators are reminded to manage the anomaly in a timely manner.

[0014] Preferably, collect multi-dimensional data from payment systems and bank reconciliation, and perform the following operations:

[0015] Collect account names, account numbers, and account types when transactions are conducted in the payment system and banking system to obtain account attribute data of the payment system and the bank;

[0016] The system collects transaction time, transaction amount, transaction summary, transaction status, counterparty information, and order number when transactions are conducted between the payment system and the banking system, thereby obtaining detailed transaction data from the payment system and the bank.

[0017] Collect data on the beginning balance, ending balance, available balance, frozen balance, interest income, interest expense, transaction fees, management fees, and fund changes of the payment system and banking system to obtain fund flow data of the payment system and banks.

[0018] Among them, multi-dimensional data for reconciliation between the payment system and the bank is determined based on the account attribute data, transaction details data and fund flow data of the payment system and the bank.

[0019] Preferably, the multi-dimensional data from the payment system and bank reconciliation are processed, and the following operations are performed:

[0020] The system cleans multi-dimensional data from payment systems and bank reconciliation, removing noisy data that is not valuable for payment reconciliation early warning, and identifies and corrects outliers in the multi-dimensional data.

[0021] Standardize the multi-dimensional data of payment system and bank reconciliation, convert the multi-dimensional data of payment system and bank reconciliation into a unified data format, remove the differences in units of measurement in the multi-dimensional data of payment system and bank reconciliation, and form standardized multi-dimensional data of payment system and bank reconciliation.

[0022] Preferably, the assessment of anomalies and transaction risk warnings in multi-dimensional data from payment systems and bank reconciliation are performed, and the following operations are carried out:

[0023] Retrieve multi-dimensional data from the user's corresponding payment system and bank reconciliation;

[0024] Retrieve all transaction counts and amounts contained in the multi-dimensional reconciliation data of the payment system and bank;

[0025] The time interval between each two adjacent transactions is obtained based on the transaction time of each transaction.

[0026] The discontinuity of a user's cash flow is obtained by using the transaction time interval between each two adjacent transactions. Where L represents the discontinuity of the user's cash flow; σ t σ represents the standard deviation of the transaction time interval; Δt represents the average time interval; σ represents the standard deviation of the transaction time interval. f Δf represents the standard deviation of transaction amount; Δf represents the average transaction amount.

[0027] Retrieve the user's average transaction frequency F;

[0028] The degree of anomaly in the multi-dimensional data of payment system and bank reconciliation is evaluated by using the discontinuity L of the fund flow corresponding to the user and the average transaction frequency F corresponding to the user, and an anomaly alarm is triggered when the degree of anomaly is large.

[0029] Preferably, the degree of anomaly in the multi-dimensional data of the payment system and bank reconciliation is assessed using the discontinuity L of the fund flow corresponding to the user and the average transaction frequency corresponding to the user, and an anomaly alarm is triggered when the degree of anomaly is large, and the following operations are performed:

[0030] Retrieve the current cash flow discontinuity L and the user's average transaction frequency at the end of each transaction.

[0031] Using the discontinuity of fund flow L as the vertical axis and the average trading frequency F corresponding to the user as the horizontal axis, the corresponding trading coordinate point (F, L) after each transaction is obtained.

[0032] Generate a transaction analysis chart based on the transaction coordinates (F, L) corresponding to the end of each transaction;

[0033] The slope between each pair of adjacent trading coordinates is obtained by analyzing the discontinuity of capital flow corresponding to each pair of adjacent trading coordinates in the trading analysis chart and the average trading frequency of the user. Where k represents the slope between any two adjacent trading coordinates; ΔL represents the difference in discontinuity of fund flow between any two adjacent trading coordinates; and ΔF represents the difference in average trading frequency between any two adjacent trading coordinates.

[0034] An anomaly index is generated by combining the slope k between each pair of adjacent transaction coordinates with the corresponding transaction coordinates (F, L) after each transaction ends.

[0035] The anomaly index is obtained using the following formula:

[0036]

[0037] Where Г represents the anomaly index; n represents the number of groups between every two adjacent transaction coordinate points; k i+1 and k i σ represents the slope of the two transaction coordinate points in the (i+1)th group and the ith group, respectively; k σ represents the standard deviation of the slope shown in the trading analysis chart. L and σ F This represents the standard deviation of the discontinuity of cash flow and the standard deviation of the average trading frequency corresponding to all trading coordinate points;

[0038] The anomaly index is compared with a preset anomaly index threshold. If the anomaly index is not lower than the preset anomaly index threshold, it indicates that the anomaly is relatively large, and an anomaly alarm is triggered.

[0039] Preferably, the multi-dimensional data from the payment system and bank reconciliation are processed, and the following operations are performed:

[0040] The payment system and bank reconciliation data are integrated into a unified data view, and the integrated unified data view of the payment system and bank reconciliation data is stored for future use.

[0041] Feature extraction is performed on multi-dimensional data of payment system and bank reconciliation. Feature vectors related to payment reconciliation early warning are extracted from the multi-dimensional data of payment system and bank reconciliation, and the feature vectors are weighted and fused to determine the feature data of payment system and bank reconciliation.

[0042] Preferably, a payment reconciliation anomaly detection model is constructed, and the following operations are performed:

[0043] Collect historical reconciliation data from payment systems and banks, and divide the collected historical reconciliation data to determine the training set and test set;

[0044] Based on machine learning technology, a training set is used to train the machine learning model, enabling the machine learning model to autonomously learn the payment reconciliation anomaly detection behavior of the payment system and the banking system from the training set, and effectively identify anomalies in payment reconciliation, thereby determining the machine learning-based payment reconciliation anomaly detection model.

[0045] The test set is input into the machine learning-based payment reconciliation anomaly detection model. The machine learning-based payment reconciliation anomaly detection model is tested according to the test set to evaluate whether the machine learning-based payment reconciliation anomaly detection model can achieve the expected effect of effectively identifying anomalies in payment reconciliation, thereby determining the model test evaluation results.

[0046] Based on the model testing and evaluation results, the parameters of the machine learning-based payment reconciliation anomaly detection model are adjusted and optimized to determine the optimal payment reconciliation anomaly detection model.

[0047] Preferably, a reconciliation report is generated based on the payment reconciliation anomaly detection results and combined with multi-dimensional data from the payment system and bank reconciliation, and the reconciliation report is output in a visual form for managers to view in real time.

[0048] According to another aspect of the present invention, a payment reconciliation early warning system based on multi-dimensional data is also provided, for implementing the payment reconciliation early warning method based on multi-dimensional data as described above, comprising:

[0049] The data acquisition module is used to collect multi-dimensional data from the payment system and bank reconciliation.

[0050] The data processing module is used to process multi-dimensional data from the payment system and bank reconciliation to determine the characteristic data of the payment system and bank reconciliation.

[0051] The anomaly detection module is used to analyze the reconciliation feature data of the payment system and the bank, and automatically identify anomalies in the payment reconciliation process of the payment system and the bank.

[0052] The early warning management module is used to automatically trigger early warnings and manage situations in a timely manner when abnormalities occur.

[0053] Compared with the prior art, the beneficial effects of the present invention are:

[0054] This invention monitors account attribute data, transaction details, and fund flow data during transactions in payment and banking systems to determine multi-dimensional data for payment system and bank reconciliation. By processing this multi-dimensional data, feature vectors are extracted and weighted to determine the characteristic data of payment system and bank reconciliation. An anomaly detection model for payment reconciliation is constructed and analyzed to automatically identify anomalies in payment reconciliation between payment and banking systems, determining the anomaly detection results. When anomalies are detected, an early warning alarm is automatically triggered, and the source of the anomaly is quickly traced, alerting administrators to manage the anomaly promptly. This invention effectively monitors payment system and bank reconciliation, promptly detects anomalies in payment reconciliation, improves reconciliation efficiency, and fully protects fund security. Attached Figure Description

[0055] Figure 1 This is a block diagram of the payment reconciliation and early warning system based on multi-dimensional data of the present invention;

[0056] Figure 2 This is a flowchart of the payment reconciliation early warning method based on multi-dimensional data according to the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] To address the current issues of ineffective monitoring of payment system and bank reconciliation, inability to promptly detect anomalies in payment reconciliation, reduced reconciliation efficiency, and insufficient guarantee of fund security, please refer to [link to relevant documentation]. Figures 1-2 This embodiment provides the following technical solution:

[0059] The payment reconciliation early warning system based on multi-dimensional data includes: a data acquisition module, a data processing module, an anomaly detection module, and an early warning management module. Specifically, through the interaction between the data acquisition module, data processing module, anomaly detection module, and early warning management module, the reconciliation between the payment system and the bank can be effectively monitored, anomalies in the payment reconciliation process can be detected in a timely manner, reconciliation efficiency can be improved, and fund security can be fully guaranteed.

[0060] The system comprises several modules: a data acquisition module for collecting multi-dimensional data from the payment system and bank reconciliation; a data processing module for processing this data and identifying characteristic features of the reconciliation process; an anomaly detection module for analyzing these characteristic data and automatically identifying anomalies in the payment and bank reconciliation processes; and an early warning management module for automatically triggering warnings and providing timely management when anomalies occur, ensuring the security of funds.

[0061] To better illustrate the payment reconciliation early warning process based on multi-dimensional data, this embodiment provides a payment reconciliation early warning method based on multi-dimensional data, implemented based on the aforementioned payment reconciliation early warning system based on multi-dimensional data, including:

[0062] Monitor transaction activity in payment and banking systems, and collect multi-dimensional data on payment system and bank reconciliation.

[0063] In this embodiment, multi-dimensional data from the payment system and bank reconciliation are collected, and the following operations are performed:

[0064] Collect account names, account numbers, and account types when transactions are conducted in the payment system and banking system to obtain account attribute data of the payment system and the bank;

[0065] It should be noted that the account name is the name of the payment system and the bank account; the account number is the account number of the bank account and the payment system account, used to uniquely identify the account; the account type is used to distinguish between corporate accounts and personal accounts, as well as the purpose of the account, such as settlement account, reserve fund account, etc.

[0066] The system collects transaction time, transaction amount, transaction summary, transaction status, counterparty information, and order number when transactions are conducted between the payment system and the banking system, thereby obtaining detailed transaction data from the payment system and the bank.

[0067] It should be noted that the transaction time is the specific time when each transaction occurs; the transaction amount includes the amount of money and the lending / borrowing direction, such as debit or credit; the transaction summary briefly describes the nature of the transaction, such as payment, refund, transfer, etc.; the transaction status includes success, failure, processing, cancellation, etc.; the counterparty information includes the counterparty's account name, account number, etc.; and the order number is the order number recorded by the payment system and the bank.

[0068] Collect data on the beginning balance, ending balance, available balance, frozen balance, interest income, interest expense, transaction fees, management fees, and fund changes of the payment system and banking system to obtain fund flow data of the payment system and banks.

[0069] It should be noted that the beginning balance is the account balance at the start of the reconciliation period; the ending balance is the account balance at the end of the reconciliation period; the available balance is the account balance that is currently available; the frozen balance is the funds that are frozen for certain reasons; and the fund change status is the inflow and outflow of each fund in the account.

[0070] Among them, multi-dimensional data for reconciliation between the payment system and the bank is determined based on the account attribute data, transaction details data and fund flow data of the payment system and the bank.

[0071] It should be noted that by monitoring account attribute data, transaction details data, and fund flow data during transactions in the payment system and banking system, multi-dimensional data for payment system and bank reconciliation can be determined, which facilitates the timely detection of anomalies in payment reconciliation between the payment system and banking system.

[0072] The multi-dimensional data of payment system and bank reconciliation are processed, feature vectors are extracted and weighted and fused to determine the feature data of payment system and bank reconciliation.

[0073] In this embodiment, the multi-dimensional data from the payment system and bank reconciliation are processed, and the following operations are performed:

[0074] The system cleans multi-dimensional data from payment systems and bank reconciliation, removing noisy data that is not valuable for payment reconciliation early warning, and identifies and corrects outliers in the multi-dimensional data.

[0075] Standardize the multi-dimensional data of payment system and bank reconciliation, convert the multi-dimensional data of payment system and bank reconciliation into a unified data format, remove the differences in the units of measurement in the multi-dimensional data of payment system and bank reconciliation, and form standardized multi-dimensional data of payment system and bank reconciliation.

[0076] The payment system and bank reconciliation data are integrated into a unified data view, and the integrated unified data view of the payment system and bank reconciliation data is stored for future use.

[0077] Feature extraction is performed on multi-dimensional data of payment system and bank reconciliation. Feature vectors related to payment reconciliation early warning are extracted from the multi-dimensional data of payment system and bank reconciliation, and the feature vectors are weighted and fused to determine the feature data of payment system and bank reconciliation.

[0078] It should be noted that by cleaning, standardizing, integrating, and extracting features from multi-dimensional data of payment systems and bank reconciliation, it is easier to analyze the feature data of payment systems and bank reconciliation and automatically identify anomalies in payment reconciliation.

[0079] Specifically, assess the degree of anomalies and issue transaction risk warnings in the multi-dimensional data of payment systems and bank reconciliation, and perform the following operations:

[0080] Retrieve multi-dimensional data from the user's corresponding payment system and bank reconciliation;

[0081] Retrieve all transaction counts and amounts contained in the multi-dimensional reconciliation data of the payment system and bank;

[0082] The time interval between each two adjacent transactions is obtained based on the transaction time of each transaction.

[0083] The discontinuity of a user's cash flow is obtained by using the transaction time interval between each two adjacent transactions. Where L represents the discontinuity of the user's cash flow; σ t σ represents the standard deviation of the transaction time interval; Δt represents the average time interval; σ represents the standard deviation of the transaction time interval. fΔf represents the standard deviation of transaction amount; Δf represents the average transaction amount.

[0084] Retrieve the user's average transaction frequency F;

[0085] The degree of anomaly in the multi-dimensional data of payment system and bank reconciliation is evaluated by using the discontinuity L of the fund flow corresponding to the user and the average transaction frequency F corresponding to the user, and an anomaly alarm is triggered when the degree of anomaly is large.

[0086] In this embodiment, the technical solution innovatively integrates the dual-dimensional fluctuation characteristics of transaction time interval and transaction amount by constructing the core quantitative indicator of fund flow discontinuity L, using the standard deviation of transaction time interval σ as the metric. t Average time interval Δt, and standard deviation of transaction amount σ f This approach breaks through the limitations of traditional single-transaction-dimensional risk identification. It transforms the complex and discrete transaction behaviors in payment systems and bank reconciliation scenarios into precisely measurable continuous risk values, enabling a quantitative and refined characterization of abnormal fluctuations in cash flow risks. This significantly improves the granularity and accuracy of risk identification, addressing the industry pain point that traditional methods struggle to capture abnormal correlations in transaction behavior. The dynamic risk monitoring and real-time early warning capabilities are built upon a time interval calculation logic driven by each transaction. In this embodiment, the above-mentioned technical solution can update data and iteratively calculate the discontinuity L of cash flow immediately after a transaction occurs, forming a dynamic analysis system combined with the average transaction frequency F. This mechanism endows the system with real-time transaction-level risk monitoring capabilities, enabling it to capture sudden changes in transaction patterns immediately. It shifts the risk identification window from "post-event batch verification" to "early warning upon transaction occurrence," gaining crucial response time for risk interception in payment systems and anomaly tracing in bank reconciliation, effectively reducing the probability of risk event propagation and handling costs. The multi-dimensional data fusion risk control system adaptability solution integrates multi-dimensional data such as transaction frequency, transaction amount, and transaction time. It takes the core perspective of "continuity of cash flow" that runs through the entire payment and reconciliation process, and covers key risk control points such as compliance of transaction behavior time patterns and verification of amount stability.

[0087] Specifically, the degree of anomaly in the multi-dimensional data of the payment system and bank reconciliation is assessed using the discontinuity L of the fund flow corresponding to the user and the average transaction frequency corresponding to the user, and an anomaly alarm is triggered when the degree of anomaly is large, and the following operations are performed:

[0088] Retrieve the current cash flow discontinuity L and the user's average transaction frequency at the end of each transaction.

[0089] Using the discontinuity of fund flow L as the vertical axis and the average trading frequency F corresponding to the user as the horizontal axis, the corresponding trading coordinate point (F, L) after each transaction is obtained.

[0090] Generate a transaction analysis chart based on the transaction coordinates (F, L) corresponding to the end of each transaction;

[0091] The slope between each pair of adjacent trading coordinates is obtained by analyzing the discontinuity of capital flow corresponding to each pair of adjacent trading coordinates in the trading analysis chart and the average trading frequency of the user. Where k represents the slope between any two adjacent trading coordinates; ΔL represents the difference in discontinuity of fund flow between any two adjacent trading coordinates; and ΔF represents the difference in average trading frequency between any two adjacent trading coordinates.

[0092] An anomaly index is generated by combining the slope k between each pair of adjacent transaction coordinates with the corresponding transaction coordinates (F, L) after each transaction ends.

[0093] The anomaly index is obtained using the following formula:

[0094]

[0095] Where Г represents the anomaly index; n represents the number of groups between every two adjacent transaction coordinate points; k i+1 and k i σ represents the slope of the two transaction coordinate points in the (i+1)th group and the ith group, respectively; k σ represents the standard deviation of the slope shown in the trading analysis chart. L and σ F This represents the standard deviation of the discontinuity of fund flows and the standard deviation of the average trading frequency corresponding to all trading coordinate points; specifically, This represents the sum of the differences between adjacent slopes, measuring the trend change magnitude of the slope sequence. When it is greater than zero, the slope continues to increase, indicating that the "sensitivity coefficient of capital flow to trading frequency" is accelerating, and the instability of the current trading pattern is continuously deteriorating. For example, a small change in trading frequency can trigger a large fluctuation in the discontinuity of capital flow, and the fluctuation amplitude is getting larger and larger. When it is not greater than zero, the slope continues to decrease, indicating that the instability of the trading pattern is converging. Used to construct a co-correction factor for "fund flow volatility" and "trading frequency volatility"; σ kIt is the standard deviation of all adjacent slopes k, measuring the dispersion of the slope sequence (i.e., the "disorderliness" of slope fluctuations), used to quantify the "disorder anomaly degree" of the slope sequence, supplementing the "dispersion anomaly" in addition to "trend anomaly" (accumulated difference). Therefore, the anomaly degree index proposed in this embodiment is used to comprehensively quantify the severity of trading anomalies from the dual dimensions of "disorder fluctuation" and "trend change," combined with compensation / penalty for fluctuations in basic indicators, effectively improving the accuracy and timeliness of trading anomaly judgment. At the same time, by judging the degree of anomaly in the above way, the judgment efficiency can be maximized while simplifying the energy consumption of capacity calculation to the greatest extent, avoiding the waste of computing power resources, and improving the accuracy and efficiency of anomaly degree judgment.

[0096] The anomaly index is compared with a preset anomaly index threshold. If the anomaly index is not lower than the preset anomaly index threshold, it indicates that the anomaly is relatively large, and an anomaly alarm is triggered.

[0097] In this embodiment, a coordinate system is constructed by integrating the discontinuity of capital flow (L) and the average transaction frequency (F). The slope reflects the changing relationship of transaction characteristics. An anomaly index Γ is then constructed by combining the slope, transaction coordinate points, and other multi-dimensional statistics (standard deviation, etc.). This comprehensively and accurately depicts abnormal transaction states from multiple dimensions, including the correlation between transaction frequency and capital flow continuity, and the pattern of characteristic fluctuations. This overcomes the limitations of single indicators or simple correlation analysis and can capture abnormal patterns in complex changes in transaction behavior. Dynamic tracking of transaction anomaly evolution is achieved by updating coordinate points, calculating the slope, and the anomaly index in real time after each transaction, enabling dynamic tracking of the development process of transaction anomalies. From immediate data after a transaction occurs, to characteristic changes between adjacent transactions, and then to the overall anomaly quantification after the accumulation of multiple sets of transactions, dynamic monitoring of the entire process of transaction anomalies from occurrence and evolution to judgment is realized, making anomaly identification more closely aligned with real-time changes in transaction scenarios. Quantifying the degree of anomaly enables intelligent early warning by converting transaction anomaly characteristics into a calculable anomaly index Γ. By comparing it with a preset threshold, quantitative judgment and intelligent early warning of the degree of anomaly are achieved. Say goodbye to vague and experience-based anomaly judgments. Based on mathematical models and statistical calculations, improve the accuracy and reliability of anomaly alerts, and provide clear and actionable quantitative standards for risk management of payment systems and bank reconciliation.

[0098] On the other hand, existing methods for identifying anomalies in financial transactions often rely on single indicators (such as exceeding limits on the amount of a single transaction or excessively high daily transaction frequency) or simple rule combinations, which are insufficient to address the complex relationships and dynamic changes in transaction behavior. The technical solution described in this embodiment integrates the continuity of capital flow, transaction frequency, and their dynamic relationship (i.e., the slope mentioned in the technical solution), as well as multi-dimensional statistical fluctuations (standard deviations), to construct a comprehensive anomaly assessment system. This system can identify more complex and concealed abnormal transaction patterns, covering anomaly scenarios corresponding to the coordinated abnormal changes in transaction frequency and discontinuity of capital flow that are easily overlooked by existing technologies. Furthermore, existing technologies mostly rely on static threshold detection or post-event batch analysis, which is insufficient for exploring dynamic changes and the correlations between indicators during the transaction process. The technical solution described in this embodiment dynamically updates data with each transaction as a node, focusing on the characteristic changes (slope) between adjacent transactions and associating the statistical characteristics of multiple sets of transactions. This strengthens the analysis of the dynamic evolution of transaction anomalies and the correlation between indicators, enabling earlier and more sensitive capture of abnormal trends and improving the timeliness and comprehensiveness of anomaly identification. Furthermore, existing technologies for anomaly detection are susceptible to subjective influence from experience thresholds. This embodiment integrates multi-dimensional trading characteristics through a mathematical model, using standard deviation to quantify data fluctuations and slope to reflect changing relationships. The constructed anomaly index Γ represents a scientific aggregation and quantification of trading anomaly characteristics. Compared to existing technologies, this reduces reliance on subjective experience, improves the scientific rigor of anomaly detection, and through multi-factor comprehensive calculation, more accurately distinguishes between normal trading fluctuations and genuine anomalies, reducing the risk of false positives and false negatives.

[0099] A payment reconciliation anomaly detection model is constructed and the reconciliation feature data of the payment system and bank are analyzed to automatically identify anomalies in the payment reconciliation of the payment system and bank system, determine the payment reconciliation anomaly detection results, and automatically trigger early warning alarms when anomalies are found.

[0100] In this embodiment, a payment reconciliation anomaly detection model is constructed, and the following operations are performed:

[0101] Collect historical reconciliation data from payment systems and banks, and divide the collected historical reconciliation data to determine the training set and test set;

[0102] Based on machine learning technology, a training set is used to train the machine learning model, enabling the machine learning model to autonomously learn the payment reconciliation anomaly detection behavior of the payment system and the banking system from the training set, and effectively identify anomalies in payment reconciliation, thereby determining the machine learning-based payment reconciliation anomaly detection model.

[0103] The test set is input into the machine learning-based payment reconciliation anomaly detection model. The machine learning-based payment reconciliation anomaly detection model is tested according to the test set to evaluate whether the machine learning-based payment reconciliation anomaly detection model can achieve the expected effect of effectively identifying anomalies in payment reconciliation, thereby determining the model test evaluation results.

[0104] Based on the model testing and evaluation results, the parameters of the machine learning-based payment reconciliation anomaly detection model are adjusted and optimized to determine the optimal payment reconciliation anomaly detection model.

[0105] In this embodiment, abnormal situations in payment reconciliation between the payment system and the banking system are automatically identified, and the following operations are performed:

[0106] Deploy the optimal payment reconciliation anomaly detection model and place it in a real payment reconciliation anomaly detection environment;

[0107] The payment system and bank reconciliation feature data are input into the payment reconciliation anomaly detection model. The model analyzes the payment system and bank reconciliation feature data and automatically identifies anomalies in the payment system and bank system during payment reconciliation, thereby determining the payment reconciliation anomaly detection result.

[0108] When anomalies are found in the payment system and banking system during payment reconciliation, an early warning alarm is automatically triggered, the source of the anomaly is quickly traced, and administrators are reminded to manage the anomaly in a timely manner.

[0109] In this embodiment, a reconciliation report is generated based on the payment reconciliation anomaly detection results and combined with multi-dimensional data from the payment system and bank reconciliation. The reconciliation report is then output in a visual format for managers to view in real time.

[0110] In summary, by monitoring account attribute data, transaction details, and fund flow data during transactions in the payment and banking systems, multi-dimensional data for payment system and bank reconciliation is determined. This data is then processed to extract feature vectors, which are then weighted and fused to identify characteristic data for payment system and bank reconciliation. By constructing a payment reconciliation anomaly detection model and analyzing this characteristic data, anomalies in payment system and bank reconciliation can be automatically identified, and the detection results can be determined. When anomalies are detected, an early warning alarm is automatically triggered, and the source of the anomaly is quickly traced, alerting administrators to manage the anomaly promptly. This approach effectively monitors payment system and bank reconciliation, promptly detects anomalies, improves reconciliation efficiency, and fully protects fund security.

[0111] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0112] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A payment reconciliation early warning method based on multi-dimensional data, characterized in that, include: Monitor transaction activity in payment and banking systems, and collect multi-dimensional reconciliation data from payment and banking systems; The multi-dimensional data of payment system and bank reconciliation are processed, feature vectors are extracted and weighted and fused to determine the feature data of payment system and bank reconciliation. Construct a payment reconciliation anomaly detection model and analyze the reconciliation feature data of the payment system and bank to automatically identify anomalies in the payment system and bank system during payment reconciliation, determine the payment reconciliation anomaly detection results, and automatically trigger early warning alarms when anomalies are found. Among them, the discontinuity L of the user's cash flow is obtained by using the transaction time interval between each two adjacent transactions, and the average transaction frequency F of the user is retrieved. The degree of anomalies in multi-dimensional data from payment systems and bank reconciliation is assessed using the discontinuity L of fund flows corresponding to users and the average transaction frequency F corresponding to users. Anomaly alerts are triggered when the degree of anomaly is significant, and the following operations are performed: Retrieve the current cash flow discontinuity L and the user's average transaction frequency at the end of each transaction. Using the discontinuity of fund flow L as the vertical axis and the average trading frequency F corresponding to the user as the horizontal axis, the corresponding trading coordinate point (F, L) after each transaction is obtained. Generate a transaction analysis chart based on the transaction coordinates (F, L) corresponding to the end of each transaction; The slope between each pair of adjacent trading coordinates is obtained by analyzing the discontinuity of capital flow corresponding to each pair of adjacent trading coordinates in the trading analysis chart and the average trading frequency of the user. Where k represents the slope between any two adjacent trading coordinates; ΔL represents the difference in discontinuity of fund flow between any two adjacent trading coordinates; and ΔF represents the difference in average trading frequency between any two adjacent trading coordinates. An anomaly index is generated by combining the slope k between each pair of adjacent transaction coordinates with the corresponding transaction coordinates (F, L) after each transaction ends. The anomaly index is obtained using the following formula: ； Where Г represents the anomaly index; n represents the number of groups between every two adjacent transaction coordinate points; k i+1 and k i σ represents the slope of the two transaction coordinate points in the (i+1)th group and the ith group, respectively; k σ represents the standard deviation of the slope shown in the trading analysis chart. L and σ F This represents the standard deviation of the discontinuity of cash flow and the standard deviation of the average trading frequency corresponding to all trading coordinate points; This represents the sum of the differences between adjacent slopes, measuring the trend change magnitude of the slope sequence; when it is greater than zero, the slope continues to increase, indicating that the "sensitivity coefficient of capital flow to trading frequency" is accelerating, and the instability of the current trading pattern is continuously deteriorating; when it is not greater than zero, the slope continues to decrease, indicating that the instability of the trading pattern is converging. Used to construct a co-correction factor for "fund flow volatility" and "trading frequency volatility"; σ k It is the standard deviation of all adjacent slopes k, which measures the dispersion of the slope sequence and is used to quantify the "disorder anomaly" of the slope sequence, supplementing the "dispersion anomaly" in addition to the "trend anomaly". The anomaly index is compared with a preset anomaly index threshold. If the anomaly index is not lower than the preset anomaly index threshold, it indicates that the anomaly is relatively large, and an anomaly alarm is triggered.

2. The payment reconciliation early warning method based on multi-dimensional data as described in claim 1, characterized in that, Assess the degree of anomalies and provide transaction risk warnings in multi-dimensional data from payment systems and bank reconciliation, and perform the following operations: Retrieve multi-dimensional data from the user's corresponding payment system and bank reconciliation; Retrieve all transaction counts and amounts contained in the multi-dimensional reconciliation data of the payment system and bank; The time interval between each two adjacent transactions is obtained based on the transaction time of each transaction. The discontinuity L of the user's cash flow is obtained by using the transaction time interval between each two adjacent transactions; Retrieve the user's average transaction frequency F; The degree of anomaly in the multi-dimensional data of payment system and bank reconciliation is evaluated by using the discontinuity L of the fund flow corresponding to the user and the average transaction frequency F corresponding to the user, and an anomaly alarm is triggered when the degree of anomaly is large.

3. The payment reconciliation early warning method based on multi-dimensional data as described in claim 2, characterized in that, Automatically identify anomalies in payment reconciliation between the payment system and the banking system, and perform the following operations: Deploy the optimal payment reconciliation anomaly detection model and place it in a real payment reconciliation anomaly detection environment; The payment system and bank reconciliation feature data are input into the payment reconciliation anomaly detection model. The model analyzes the payment system and bank reconciliation feature data and automatically identifies anomalies in the payment system and bank system during payment reconciliation, thereby determining the payment reconciliation anomaly detection result. When anomalies are found in the payment system and banking system during payment reconciliation, an early warning alarm is automatically triggered, the source of the anomaly is quickly traced, and administrators are reminded to manage the anomaly in a timely manner.

4. The payment reconciliation early warning method based on multi-dimensional data as described in claim 1, characterized in that, Collect multi-dimensional data from payment systems and bank reconciliation, and perform the following operations: Collect account names, account numbers, and account types when transactions are conducted in the payment system and banking system to obtain account attribute data of the payment system and the bank; The system collects transaction time, transaction amount, transaction summary, transaction status, counterparty information, and order number when transactions are conducted between the payment system and the banking system, thereby obtaining detailed transaction data from the payment system and the bank. Collect data on the beginning balance, ending balance, available balance, frozen balance, interest income, interest expense, transaction fees, management fees, and fund changes of the payment system and banking system to obtain fund flow data of the payment system and banks. Among them, multi-dimensional data for reconciliation between the payment system and the bank is determined based on the account attribute data, transaction details data and fund flow data of the payment system and the bank.

5. The payment reconciliation early warning method based on multi-dimensional data as described in claim 4, characterized in that, Process multi-dimensional data from payment systems and bank reconciliation, and perform the following operations: The system cleans multi-dimensional data from payment systems and bank reconciliation, removing noisy data that is not valuable for payment reconciliation early warning, and identifies and corrects outliers in the multi-dimensional data. Standardize the multi-dimensional data of payment system and bank reconciliation, convert the multi-dimensional data of payment system and bank reconciliation into a unified data format, remove the differences in units of measurement in the multi-dimensional data of payment system and bank reconciliation, and form standardized multi-dimensional data of payment system and bank reconciliation.

6. The payment reconciliation early warning method based on multi-dimensional data as described in claim 5, characterized in that, Process multi-dimensional data from payment systems and bank reconciliation, and perform the following operations: The payment system and bank reconciliation data are integrated into a unified data view, and the integrated unified data view of the payment system and bank reconciliation data is stored for future use. Feature extraction is performed on multi-dimensional data of payment system and bank reconciliation. Feature vectors related to payment reconciliation early warning are extracted from the multi-dimensional data of payment system and bank reconciliation, and the feature vectors are weighted and fused to determine the feature data of payment system and bank reconciliation.

7. The payment reconciliation early warning method based on multi-dimensional data as described in claim 6, characterized in that, Build a payment reconciliation anomaly detection model and perform the following operations: Collect historical reconciliation data from payment systems and banks, and divide the collected historical reconciliation data to determine the training set and test set; Based on machine learning technology, a training set is used to train the machine learning model, enabling the machine learning model to autonomously learn the payment reconciliation anomaly detection behavior of the payment system and the banking system from the training set, and effectively identify anomalies in payment reconciliation, thereby determining the machine learning-based payment reconciliation anomaly detection model. The test set is input into the machine learning-based payment reconciliation anomaly detection model. The machine learning-based payment reconciliation anomaly detection model is tested according to the test set to evaluate whether the machine learning-based payment reconciliation anomaly detection model can achieve the expected effect of effectively identifying anomalies in payment reconciliation, thereby determining the model test evaluation results. Based on the model testing and evaluation results, the parameters of the machine learning-based payment reconciliation anomaly detection model are adjusted and optimized to determine the optimal payment reconciliation anomaly detection model.

8. The payment reconciliation early warning method based on multi-dimensional data as described in claim 7, characterized in that, Based on the results of payment reconciliation anomaly detection and combined with multi-dimensional data from the payment system and bank reconciliation, a reconciliation report is generated and output in a visual format for managers to view in real time.

9. A payment reconciliation early warning system based on multi-dimensional data, used to implement the payment reconciliation early warning method based on multi-dimensional data as described in claim 8, characterized in that, include: The data acquisition module is used to collect multi-dimensional data from the payment system and bank reconciliation. The data processing module is used to process multi-dimensional data from the payment system and bank reconciliation to determine the characteristic data of the payment system and bank reconciliation. The anomaly detection module is used to analyze the reconciliation feature data of the payment system and the bank, and automatically identify anomalies in the payment reconciliation process of the payment system and the bank. The early warning management module is used to automatically trigger early warnings and manage situations in a timely manner when abnormalities occur.

Images