A method and system for checking the balance of stored-value card for ETC
Through deep learning and multi-model detection methods, a stored value card balance audit system was designed, which solved the problems of complex balance calculation and large data analysis in the stored value card system, realized abnormal positioning and capital risk control, and improved operational efficiency and Data analysis capabilities.
Patent Information
- Application Number
- CN202411045820.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-08-01
AI Technical Summary
In the toll-connected toll operation of the existing stored value card system, software systems involving multiple developers have complicated balance calculations, resulting in endless balance abnormalities, and production systems cannot support the unified analysis and calculation of large data volumes throughout the life cycle. , lack of algorithm models for review and analysis.
Using innovative methods of deep learning and multi-model detection, we design ETC stored value card balance auditing methods and systems, and obtain original data and generate ledgers through single card and batch audit models, analyze exception details, and perform data cleaning and exception processing. Generate an audit report to achieve accurate calculation and abnormal positioning of system balances.
Effectively analyze and locate abnormal balance of stored value card, improve the financial risk control capabilities of the operation management party, solve the problem of unified analysis of large data volume, and have the ability to review and analyze the balance of the system, which improves efficiency and accuracy.
Smart Images

Figure CN118586937B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of electronic information and data processing technology, and in particular relates to a method and system for checking the balance of an ETC stored-value card. Background Art
[0002] Stored-value card users, highway operation managers, and extended application managers perceive the card balance. Recharge, toll consumption, and extended consumption services complete related operations by reading the card balance in real time. The stored-value card issuer and operation manager manage the system balance based on the card balance. The system balance is calculated based on a series of abnormal judgments, accounting, splitting, and settlement operations for all stored-value card services. It reflects the real business of the stored-value card and the real stored-value card balance. The card balance should be consistent with the system balance.
[0003] The use of existing stored-value cards still has the following problems:
[0004] 1. In the networked toll collection operation and service system of toll roads, there are many entities involved in the calculation of the system balance and card balance of stored-value cards. There are more than 10 major software systems from different developers in Guangdong Province alone. In addition, the links and time spans involved in the calculation are also very long, and various abnormal balance problems emerge in an endless stream. Analysis and positioning require huge manpower, time and computing resources. At present, there is no algorithm model in the industry that can analyze the abnormalities of the system balance and card balance throughout the life cycle of stored-value cards, locate the nodes where abnormalities occur, summarize the abnormalities of stored-value cards from the perspective of a single card, and perceive the financial risks of stored-value cards from the perspective of operations and management.
[0005] 2. The life cycle of stored-value cards is long. Guangdong Province's national standard stored-value cards have been in use for nearly 10 years, and the amount of business data generated during this period is very large. According to statistics, there are more than 1 billion active stored-value card business flows throughout their life cycle, and the number of full-volume stored-value cards can reach tens of billions. The data is stored in a decentralized manner and is still increasing. The existing production system cannot support the unified analysis and calculation of such a large amount of data.
[0006] 3. The system balance reflects the real stored-value card balance. Ensuring that the system balance is correctly calculated is the focus of the production system. The existing production system can only meet the construction and maintenance of stored-value card application functions, and does not have an algorithm model for review and analysis.
[0007] In summary, the study of algorithm models with stored-value card balance auditing function is of great significance to all participants in the toll road network toll operation and service system. Summary of the invention
[0008] The purpose of the present invention is to design a method and system for auditing the balance of ETC stored-value cards, which integrates the innovations of deep learning and multi-model detection, and realizes the audit of the ETC stored-value card system balance and the card balance. The stored-value cards in the offline transaction mode can be audited and exceptions can be handled through the algorithm model.
[0009] In order to achieve the above object, a first aspect of the present invention provides a method for checking the balance of an ETC stored-value card, the method comprising:
[0010] Step 1: Obtain the original data of the high-speed stored-value card and generate a stored-value card account; wherein the original data is the original detailed business data of all parties in the entire life cycle of the stored-value card;
[0011] Step 2: Build a single card audit model to analyze the original data to obtain the abnormal details of the stored-value card; the abnormal details include the stored-value card number, business time, abnormal type, abnormal amount and remarks. The abnormal type and abnormal amount are summarized to form an analysis report of the stored-value card. At the same time, the abnormal balance information is summarized and counted according to the abnormal type, including:
[0012] Step 2.1, initialize the stored-value card ledger data and add auxiliary analysis columns for analysis;
[0013] Step 2.2, further analyze the suspected abnormal location data of the stored-value card number, the starting sequence number and the ending sequence number to determine whether there is any abnormality in the data;
[0014] Step 2.3, cleaning the suspected abnormal location data to obtain abnormal location data;
[0015] Step 2.4: locate the abnormal flow according to the abnormal location data, and locate the abnormal features in the remarks;
[0016] Step 2.5: Output the abnormal details in the form of an analysis report of a single stored-value card and a summary of the abnormal balance information;
[0017] Step 3: Aggregate the original data to obtain aggregated data; wherein the aggregated data includes aggregated information and detailed information, the aggregated information includes the original aggregated business data during the stored-value card audit period and the aggregated data of the stored-value card system balance change record table and the beginning and ending system balance snapshots, and the detailed information includes the audit period, stored-value card number, settlement amount, recharge amount, refund amount, fund transfer amount, complaint refund amount and received balance;
[0018] Step 4: Build a batch audit model to analyze the summary data to obtain a summary table, and generate a monthly audit report for stored-value cards based on the summary table, including:
[0019] Step 4.1, using the summary data as input, simulating the production system to calculate the system balance, determining the difference category and locating the problematic stored-value card by comparing the system balance;
[0020] Step 4.2, the detailed data of the problem stored-value cards are compared by common identification, and the records of differences are generated to form a stored-value card batch audit difference summary table, and the stored-value card batch audit difference detailed table is matched according to the stored-value card batch audit difference summary table;
[0021] Step 4.3: Generate a monthly audit report based on the stored-value card batch audit difference summary table and the stored-value card batch audit difference detail table.
[0022] Furthermore, the step 2 further comprises the following steps:
[0023] Step 2.6: Use reinforcement learning to process the abnormal details generated by the single card audit model, including:
[0024] The feature vector of the abnormal details information generated by the single card audit model is used as the state , define the action set A and reward function ;
[0025] The objective function is designed by combining the deterministic policy gradient method and the context-aware decision tree, which is expressed as follows:
[0026] ,
[0027] in, Represents the policy parameters; Represents the policy function, which consists of parameters definition; According to the behavior strategy The resulting state distribution; Indicates in status Next Action Instant rewards received; is the value function of the next state; A coefficient that represents the balance between immediate rewards and long-term value;
[0028] The hybrid function is designed to optimize both the performance and stability of the strategy, as shown below:
[0029] ,
[0030] in, Indicates the strategy Divergence, ensuring that the new strategy does not deviate too far from the old strategy, maintaining the stability of the learning process; represents the regularization parameter, controlling The impact of divergence on the total loss; then use gradient ascent to update the policy parameters, expressed as follows:
[0031] ,
[0032] in, is the learning rate;
[0033] Using a deterministic strategy Directly from the current state Generate actions.
[0034] Furthermore, the state Including at least one of the transaction amount, the difference from the previous transaction amount, the time interval of the previous transaction, the transaction time of the day, the abnormal frequency in the account history and the account level; the action set The method includes at least one of automatically adjusting the account balance, sending an alert to the user to request confirmation, and forwarding the request to manual review; wherein the reward function is designed as follows:
[0035] , when the action correctly resolves the exception and minimizes user disruption;
[0036] , when the action fails to resolve the exception or adds burden to the user;
[0037] , when no further action is required.
[0038] Furthermore, the ledger includes basic information, access information and card balance information; wherein, the basic information includes at least one of the stored-value card number C, business time t, business type B and business status S; the access information includes at least one of the access serial number, exit license plate, vehicle type, entry station, entry time Te, clearing time Tc and transaction location / exit station; the card balance information includes at least one of the receivable amount Ap, discount amount Ad, transaction amount Af, card balance before transaction Ab, card deduction amount At and card balance after transaction Aa.
[0039] Furthermore, the auxiliary analysis column includes:
[0040] Calculate whether the first difference and the second difference are 0 within the transaction time T. 0 indicates that the transaction is normal, otherwise it indicates that the transaction is abnormal. The first difference is the difference between the balance and the transaction amount before the card transaction and the balance after the card transaction, and the second difference is the difference between the transaction amount and the next transaction record.
[0041] Initialize the balance before and after the suggested card transaction;
[0042] Calculate whether the first modified difference and the second modified difference are 0 within the transaction time T, 0 indicates normal transaction, otherwise the transaction is abnormal; among which, the first difference is the difference between the balance and transaction amount before the recommended card transaction and the balance after the recommended card transaction, and the second difference is the difference between the balance after the recommended card transaction and the next transaction record.
[0043] Further, when the first modified difference and the second modified difference are both 0, further check the suspected abnormal location data, wherein the suspected abnormal location data includes at least one of the following:
[0044] A. If the difference after the first modification and the difference after the second modification show the characteristics of step-like opposite numbers, it is considered that the lane software has filled in the balance before the card transaction incorrectly in the uploaded flow;
[0045] B. The Lane software incorrectly fills in the balance before / after the card transaction in the uploaded flow;
[0046] C. The transaction date filled in incorrectly in the uploaded transaction records by the Lane Software;
[0047] D. The Lane software entered the wrong amount in the uploaded transaction flow.
[0048] Furthermore, the data cleaning includes at least one of the following:
[0049] a. If the difference between the first modification and the second modification presents a step-by-step reverse number of transactions, the transaction amount before and after the last line of abnormal transactions shall be used as the basis, and the data cleaning method of reverse chronological order shall be used for cleaning;
[0050] b. If the difference between the first modification and the second modification is 0, the suspected abnormal transaction will be cleaned using the data cleaning method of positive time sequence based on the balance before the card transaction in the first line of abnormal transaction;
[0051] c. For suspected abnormal transactions with incorrect transaction time, if the entry time and transaction time T are close, the time will be calculated based on the clearing time - 1.
[0052] Furthermore, the abnormal characteristics include at least one of the following: failure to deduct fees from the current transaction, failure to deduct fees from the upper and lower transactions, missing transactions, the first half of the recharge, and erroneous push of card limit adjustment data.
[0053] Furthermore, the step 4 also includes, step 4.4, processing the summary difference data output by the batch audit model, including:
[0054] The identified abnormal data is used to perform anomaly analysis based on the gradient boosting tree as follows:
[0055] ,
[0056] in, represents the input features, Indicates the exception category, represents the model parameters;
[0057] The decision-making algorithm defines the processing strategy according to the exception category, which is expressed as follows:
[0058] ,
[0059] in, Indicates the determination of specific processing actions; represents the decision tree parameters;
[0060] The continuous learning mechanism optimizes decision accuracy and processing efficiency, setting the optimization goals as:
[0061] ,
[0062] in, Indicates the actual processing result. represents the loss function, which is used to evaluate the deviation between the prediction strategy and the actual result. represents the i-th input feature.
[0063] In a second aspect of the present invention, a system for checking the balance of an ETC stored-value card is provided, the system comprising:
[0064] A single card audit model is used to obtain the original data of a high-speed stored-value card and generate a stored-value card ledger; wherein the original data is the original detailed business data of all parties in the entire life cycle of a single stored-value card; a single card audit model is constructed to analyze the original data to obtain abnormal detailed information of a single stored-value card; wherein the abnormal details include the stored-value card number, business time, abnormal type, abnormal amount and remarks, wherein the abnormal type and abnormal amount are summarized to form an analysis report of the single stored-value card, and the abnormal balance information is summarized and counted according to the abnormal type, including:
[0065] Step 2.1, initialize the stored-value card ledger data and add auxiliary analysis columns for analysis;
[0066] Step 2.2, further analyze the suspected abnormal location data of the stored-value card number, the starting sequence number and the ending sequence number to determine whether there is any abnormality in the data;
[0067] Step 2.3, cleaning the suspected abnormal location data to obtain abnormal location data;
[0068] Step 2.4: locate the abnormal flow according to the abnormal location data, and locate the abnormal features in the remarks;
[0069] Step 2.5: Output the abnormal details in the form of an analysis report of a single stored-value card and a summary of the abnormal balance information;
[0070] The batch audit model is used to aggregate the original data to obtain the aggregated data; wherein the aggregated data includes aggregated information and detailed information, the aggregated information includes the original aggregated business data during the stored-value card audit period and the aggregated data of the stored-value card system balance change record table and the snapshots of the system balance at the beginning and end of the period, and the detailed information includes the audit period, stored-value card number, settlement amount, recharge amount, refund amount, fund transfer amount, complaint refund amount and received balance; the batch audit model is constructed to analyze the aggregated data to obtain the aggregated table, and the monthly audit report of the stored-value card is generated according to the aggregated table, including:
[0071] Step 4.1, using the summary data as input, simulating the production system to calculate the system balance, determining the difference category and locating the problematic stored-value card by comparing the system balance;
[0072] Step 4.2, the detailed data of the problem stored-value cards are compared by common identification, and the records of differences are generated to form a stored-value card batch audit difference summary table, and the stored-value card batch audit difference detailed table is matched according to the stored-value card batch audit difference summary table;
[0073] Step 4.3: Generate a monthly audit report based on the stored-value card batch audit difference summary table and the stored-value card batch audit difference detail table.
[0074] The beneficial technical effects of the present invention are at least as follows:
[0075] (1) The single-card audit model of the stored-value card in the present invention can effectively analyze the endless problems of stored-value card balances caused by a large number of participants. It can accurately locate and analyze abnormal stored-value card balances, transforming the original purely manual query and analysis into a systematic one through an algorithm model, greatly improving efficiency. For the stored-value card operation and management party, it can improve the ability to control capital risks, and for the stored-value card users, it can prevent the interests from being damaged.
[0076] (2) The single card and batch audit model in the present invention is based on cloud platform big data deployment, development and analysis, which can solve the problem that the amount of stored-value card business data is huge and the production system cannot support the unified analysis and calculation of the large amount of data.
[0077] (3) The batch audit model in the present invention has the system balance review and analysis capabilities that the production system does not have. For stored-value cards with abnormalities, the detailed records of the abnormalities can be accurately located.
[0078] (4) Machine learning-based methods are very suitable for batch audits involving large and complex data volumes, covering multiple accounts and their transactions, and quickly identifying patterns and anomalies from large amounts of data. Batch audits often involve different types of anomalies, which may involve complex cross-dependencies and patterns, not just single account issues. BRIEF DESCRIPTION OF THE DRAWINGS
[0079] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.
[0080] Figure 1 The present invention is a flow chart of a method for checking the balance of an ETC stored-value card according to an embodiment of the present invention.
[0081] Figure 2 This is a schematic diagram of an error in filling in the balance before a card transaction in the uploaded flow by the lane software in an embodiment of the present invention.
[0082] Figure 3 This is a schematic diagram of an error in filling in the balance before / after a card transaction in the uploaded flow by the Lane software in an embodiment of the present invention.
[0083] Figure 4 This is a schematic diagram of an error in filling in the transaction date in the uploaded transaction flow by the Lane software in an embodiment of the present invention.
[0084] Figure 5 This is a schematic diagram of an error filled in by the lane software in the uploaded flow according to an embodiment of the present invention.
[0085] Figure 6 This is a schematic diagram of a fee deduction failure (this flow) in an embodiment of the present invention.
[0086] Figure 7 This is a schematic diagram of fee deduction failure (up and down flow) in an embodiment of the present invention.
[0087] Figure 8 Schematic diagram of water flow loss in an embodiment of the present invention.
[0088] Fig. 9 This is a schematic diagram of the first half of recharging according to an embodiment of the present invention.
[0089] Fig.10 This is a schematic diagram of erroneously pushing card limit adjustment data according to an embodiment of the present invention.
[0090] Fig.11 Schematic diagram of matching the summary data of the embodiment of the present invention with the same category of data in the system balance change table Figure 1 .
[0091] Fig.12Schematic diagram of matching the summary data of the embodiment of the present invention with the same category of data in the system balance change table Figure 2 .
[0092] Fig.13 Schematic diagram of matching the summary data of the embodiment of the present invention with the same category of data in the system balance change table Figure 3 .
[0093] Fig.14 Schematic diagram of matching the summary data of the embodiment of the present invention with the same category of data in the system balance change table Figure 4 .
[0094] Fig.15 Schematic diagram of matching the summary data of the embodiment of the present invention with the same category of data in the system balance change table Figure 5 .
[0095] Fig.16 It is a schematic diagram for accurately recording the differences between the embodiments of the present invention.
[0096] Fig.17 This is a flowchart of abnormal analysis of stored-value card balance in the Huawei Cloud Data Lake Governance Center in an embodiment of the present invention.
[0097] Fig.18 This is a framework diagram of the ETC stored-value card balance audit system according to an embodiment of the present invention. DETAILED DESCRIPTION
[0098] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.
[0099] In one or more embodiments, Figure 1 As shown, a method for checking the balance of an ETC stored-value card is disclosed. The method includes steps one to four, including:
[0100] Step 1: Obtain the original data of the high-speed stored-value card and generate a stored-value card ledger; wherein the original data is the original detailed-level business data of all parties in the entire life cycle of a single stored-value card.
[0101] Specifically, the ledger includes basic information, access information and card balance information; wherein, the basic information includes at least one of the stored-value card number C, business time t, business type B and business status S; the access information includes at least one of the access serial number, exit license plate, vehicle type, entry station, entry time Te, clearing time Tc and transaction location / exit station; the card balance information includes at least one of the receivable amount Ap, discount amount Ad, transaction amount Af, card balance before transaction Ab, card deduction amount At and card balance after transaction Aa.
[0102] Step 2: Build a single card audit model to analyze the original data to obtain the abnormal details of the stored-value card; the abnormal details include the stored-value card number, business time, abnormal type, abnormal amount and remarks. The abnormal type and abnormal amount are summarized to form an analysis report of the stored-value card. At the same time, the abnormal balance information is summarized and counted according to the abnormal type, including:
[0103] Step 2.1, initialize the stored-value card ledger data and add auxiliary analysis columns for analysis;
[0104] Specifically, the auxiliary analysis column includes:
[0105] Calculate whether the first difference and the second difference are 0 within the transaction time T. 0 indicates that the transaction is normal, otherwise it indicates that the transaction is abnormal. The first difference is the difference between the balance and the transaction amount before the card transaction and the balance after the card transaction, and the second difference is the difference between the transaction amount and the next transaction record.
[0106] Initialize the balance before and after the suggested card transaction;
[0107] Calculate whether the first modified difference and the second modified difference are 0 within the transaction time T, 0 indicates normal transaction, otherwise the transaction is abnormal; among which, the first difference is the difference between the balance and transaction amount before the recommended card transaction and the balance after the recommended card transaction, and the second difference is the difference between the balance after the recommended card transaction and the next transaction record.
[0108] When the difference after the first modification and the difference after the second modification are both 0, further check the suspected abnormal location data, where the suspected abnormal location data includes at least one of the following:
[0109] A. If the difference after the first modification and the difference after the second modification show the characteristics of step-like opposite numbers, it is considered that the lane software has filled in the balance before the card transaction incorrectly in the uploaded flow;
[0110] B. The Lane software incorrectly fills in the balance before / after the card transaction in the uploaded flow;
[0111] C. The transaction date filled in incorrectly in the uploaded transaction records by the Lane Software;
[0112] D. The Lane software entered the wrong amount in the uploaded transaction flow.
[0113] Specifically, Figure 2-5 As shown, suspected abnormal positioning refers to the information such as the balance before and after the card transaction and the transaction time in the toll-free section generated by the toll-free section. The auxiliary analysis column is used to locate the abnormality. The positioning content is the stored-value card number, the starting sequence number and the ending sequence number, which is convenient for subsequent data cleaning. The premise of abnormal interval positioning is that at least one or more of the flow before and after the interval is a normal flow, that is, the modified difference 1Adif1 and the modified difference 2Adif2 of the flow are both 0. The following is the specific content of locating suspected abnormalities:
[0114] (1) The modified difference 1 Adif1 and the modified difference 2 Adif2 showed the characteristics of step-by-step opposite numbers. The reason was that the lane software filled in the balance before the card transaction incorrectly in the uploaded flow, and it was actually determined to be normal;
[0115] (2) In addition to point (1), regarding the exception that the total of modified difference 1Adif1 and modified difference 2Adif2 is 0, including the total of modified difference 1Adif1 and modified difference 2Adif2 of a single or multiple records is 0, the reason is that the lane software incorrectly fills in the balance before / after the card transaction in the uploaded flow, and is actually determined to be no exception;
[0116] (3) The transaction time is greater than the clearing date (settlement date) because the Lane software incorrectly fills in the transaction date when uploading the transaction, resulting in the transaction appearing in the wrong position and the calculation errors of the modified difference 1Adif1 and the modified difference 2Adif2.
[0117] (4) The balances Aaa and Aab before and after the card transaction suddenly jumped to abnormally large values. The reason was that the lane software filled in the transaction incorrectly when uploading it.
[0118] Step 2.2, further analyze the suspected abnormal location data of the stored-value card number, the starting sequence number and the ending sequence number to determine whether there is any abnormality in the data;
[0119] Step 2.3, cleaning the suspected abnormal location data to obtain abnormal location data;
[0120] Preferably, the data cleaning includes at least one of the following:
[0121] a. If the difference between the first modification and the second modification presents a step-by-step reverse number of transactions, the transaction amount before and after the last line of abnormal transactions shall be used as the basis, and the data cleaning method of reverse chronological order shall be used for cleaning;
[0122] b. If the difference between the first modification and the second modification is 0, the suspected abnormal transaction will be cleaned using the data cleaning method of positive time sequence based on the balance before the card transaction in the first line of abnormal transaction;
[0123] c. For suspected abnormal transactions with incorrect transaction time, if the entry time and transaction time T are close, the time will be calculated based on the clearing time - 1.
[0124] Specifically, data cleaning refers to making corrections based on the original data. The corrections include the balances before and after the proposed card transaction (Aaa, Aab), the difference 1 (Adif1) after the modification, the difference 2 (Adif2) after the modification, and the transaction time. The purpose of data cleaning is to exclude the above-mentioned suspected abnormalities but no abnormalities in reality, so as to facilitate the subsequent analysis of abnormal balances of stored-value cards. There are three types of data cleaning:
[0125] (1) For the suspected anomaly point (1), the modified difference 1 Adif1 and the modified difference 2 Adif2 present a step-by-step reverse number of transactions. The balance before and after the card transaction in the last line of abnormal transactions is taken as the basis. The data is cleaned in reverse chronological order. The cleaning formula is "Aab = Aaa - Af", "Pre (Aaa) = Aab", Pre is the previous record, and the cleaning field is the balance Ab before the card transaction.
[0126] (2) For other suspected abnormal transactions where the sum of difference 1 and difference 2 is 0, the balance before the card transaction in the first line of abnormal transaction is used as the basis and the data is cleaned in chronological order. The cleaning formula is "Aaa=Aab+Af", "Next(Aab)=Aaa", and the cleaning fields are the balances before and after the card transaction, Aaa and Aab.
[0127] (3) For suspected abnormal transactions with incorrect transaction time, if the entry time and transaction time T are close, the clearing time shall be used as the basis, and the cleaning formula is "T=Tc-1"; otherwise, the cleaning formula is "T=Te+(Tc-Te) / 2".
[0128] After cleaning, the modified difference 1Adif1 and the modified difference 2Adif2 need to be recalculated according to the initialization formula. If there are multiple suspected anomalies, multiple cleanings need to be performed until all suspected anomalies disappear.
[0129] Step 2.4: Locate the abnormal flow based on the abnormal location data, and locate the abnormal features in the remarks.
[0130] Specifically, the abnormal characteristics include at least one of the following: failure to deduct fees from the current transaction, failure to deduct fees from the upper and lower transactions, missing transactions, the first half of the recharge, and erroneous push of card limit adjustment data.
[0131] Specifically, after data cleaning is completed, the abnormal flow can be located according to the abnormal balance characteristics. At the same time, the abnormal time, abnormal type and amount are recorded in the notes for easy verification. Some abnormal balance locations require the association of stored-value card related data outside the ledger. As shown in 6-10, the specific abnormal characteristics are as follows:
[0132] (1) Deduction failed (this transaction). Logical judgment formula: Adif1<0andAt=0. Analyzing this exception can effectively improve the capital risk control capability of the stored-value card operation and management party.
[0133] (2) Deduction failure (up and down flow). Logical judgment formula: Adif2<0andNext(Aab)=Pre(Aaa). Analyzing this exception can effectively improve the capital risk control ability of the stored-value card operation and management party.
[0134] (3) Transaction missing. Logical judgment formula: Adif2<0. Analyzing this anomaly can reveal whether the rights and interests of stored-value card users are damaged.
[0135] (4) The first half of the recharge. Logical judgment formula: Adif2<0andNext(Aab)<>Pre(Aab), and there is an unsuccessful recharge record in the recharge table before the next transaction time T, with the difference 2Adif2 amount opposite to the modified amount. Analyzing this anomaly can effectively improve the capital risk control ability of the stored-value card operation and management party.
[0136] (4) The first half of the recharge. Logical judgment formula: Adif2<0andNext(Aab)<>Pre(Aab), and there is an unsuccessful recharge record in the recharge table before the next transaction time T, with the difference 2Adif2 amount opposite to the modified amount. Analyzing this anomaly can effectively improve the capital risk control ability of the stored-value card operation and management party.
[0137] (5) Card limit adjustment data is pushed incorrectly. Logical judgment formula: Af=0andAt<>0andAdif1=0andAdif2=0. Analyzing this anomaly can effectively improve the capital risk control capability of the stored-value card operation and management party.
[0138] Step 2.5: Output the abnormal details in the form of an analysis report of a single stored-value card and a summary statistical balance abnormality information.
[0139] Specifically, after completing the location and analysis of the balance anomaly, various detailed anomaly information is collected based on the dimension of the stored-value card. The anomaly details include the stored-value card number, business time, anomaly type, anomaly amount, and remarks. The anomaly type and anomaly amount are summarized to form an analysis report for a single stored-value card, which is one of the output forms. Another output form is to summarize and count the balance anomaly information based on the balance anomaly type for overall and special analysis. The summarized balance anomaly information includes the anomaly type, time period, quantity, and anomaly amount.
[0140] Step 2.6: Use reinforcement learning to process the abnormal details generated by the single card audit model, including:
[0141] The feature vector of the abnormal details information generated by the single card audit model is used as the state , define a set of actions And the reward function ;
[0142] The objective function is designed by combining the deterministic policy gradient method and the context-aware decision tree, which is expressed as follows:
[0143] ,
[0144] in, Represents the policy parameters; Represents the policy function, which consists of parameters definition; According to the behavior strategy The resulting state distribution; Indicates in status Next Action Instant rewards received; is the value function of the next state; A coefficient that represents the balance between immediate rewards and long-term value;
[0145] The hybrid function is designed to optimize both the performance and stability of the strategy, as shown below:
[0146] ,
[0147] in, Indicates the strategy Divergence, ensuring that the new strategy does not deviate too far from the old strategy, maintaining the stability of the learning process; represents the regularization parameter, controlling The impact of divergence on the total loss; then use gradient ascent to update the policy parameters, expressed as follows:
[0148] ,
[0149] in, is the learning rate;
[0150] Using a deterministic strategy Directly from the current state Generate actions.
[0151] Preferably, the state Including at least one of the transaction amount, the difference from the previous transaction amount, the time interval of the previous transaction, the transaction time of the day, the abnormal frequency in the account history and the account level; the action set The method includes at least one of automatically adjusting the account balance, sending an alert to the user to request confirmation, and forwarding the request to manual review; wherein the reward function is designed as follows:
[0152] , when the action correctly resolves the exception and minimizes user disruption;
[0153] , when the action fails to resolve the exception or adds burden to the user;
[0154] , when no further action is required.
[0155] Preferably, the input part of the single card audit model is to extract the original detailed business data of all parties in the entire life cycle of the stored-value card as the input part of the algorithm model, including:
[0156] (1) Recharge, refund and transfer data generated by the stored-value card operation and management party;
[0157] (2) The original transaction data of traffic and expansion transactions of the road section owners and expansion application managers;
[0158] (3) Stored-value card settlement data and abnormality determination data of the provincial networking center.
[0159] The data is used to generate a stored-value card ledger, which is the basis for abnormal analysis of stored-value card balances. The ledger includes three types of information: basic information, access information, and card balance information. The details are as follows:
[0160] (1) D basic information: including stored-value card number C, service time t, service type (pass, recharge, etc.) B, service status S;
[0161] (2) D. Passage information: mainly includes the pass serial number (unique identifier), exit license plate, vehicle type, entry station, entry time Te, clearing time Tc, transaction location / exit station, etc.;
[0162] (3) D Card balance information: including receivable amount Ap, discount amount Ad, transaction amount Af, card balance before transaction Ab (the balance on the card read before the transaction), card deduction amount At, and card balance after transaction Aa (the balance on the card read after the transaction).
[0163] Specifically, add an auxiliary analysis column based on the ledger and initialize it before analysis. The auxiliary analysis column is a tool for analyzing abnormal use of stored-value card balances. The specific content and uses include:
[0164] Sequence number Seq: Initialization formula: seq=SORT(t), each card is sorted starting from 1 according to the business time (exit time for passing and recharge confirmation time for recharge).
[0165] Purpose: To establish a stored-value card ledger, to connect the upper and lower transaction records of the same card, to facilitate analysis of abnormal balances.
[0166] Difference 1 Dif1: Initialization formula: Dif1=Ab+Af-Aa;
[0167] Purpose: If the transaction is normal, it will be changed to 0 to judge the abnormal balance; display the original data.
[0168] Difference 2 Dif2: Initialization formula: Dif2=Af-Next(Ab), Next is the next record;
[0169] Purpose: If the transaction is normal, it will be changed to 0 to judge the abnormal balance; display the original data.
[0170] The balance before the recommended card transaction is Aab: Initialization formula: Aab=Ab;
[0171] Purpose: Used to display corrected data after data cleaning.
[0172] Difference after modification 1Adif1: Initialization formula, Adif1=Aab+Af-Aaa;
[0173] Purpose: If the transaction is normal, it will be changed to 0 to judge the abnormal balance; after data cleaning, the corrected data will be displayed.
[0174] Modified difference 2Adif2: Initialization formula, Adif2=Aaa-Next(Aab), Next is the next record;
[0175] Purpose: If the transaction is normal, it will be changed to 0 to judge the abnormal balance; after data cleaning, the corrected data will be displayed.
[0176] Recommended trading time T: Initialization formula, T=t;
[0177] Purpose: Used to determine whether the transaction time is incorrect; display the corrected data after data cleaning.
[0178] Step three: Summarize the original data to obtain summary data; wherein the summary data includes summary information and detailed information, the summary information includes the original summary business data during the stored-value card audit period, and the summary data of the stored-value card system balance change record table and the beginning and ending system balance snapshots, and the detailed information includes the audit period, stored-value card number, settlement amount, recharge amount, refund amount, fund transfer amount, complaint refund amount and received balance.
[0179] Specifically, the model is mainly used to verify whether the stored-value card system balance of the production system is calculated according to the design, and when there is a deviation, locate the node where the abnormality occurs. The model includes three parts: input, calculation and analysis, and output, as follows:
[0180] The overall input is divided into two parts. The first part is the summary information, represented as DSummary, including the original summary business data DBusinessSummary during the stored-value card audit period and the summary data DBalanceChangeSummary of the stored-value card system balance change record table, as well as the beginning and end system balance snapshots SBeginning of Period and SEnd of Period, which are used to locate the nodes where the stored-value card system balance calculation deviation occurs. The variable set in DBusinessSummary data includes the audit period T, stored-value card number C, settlement amount As, recharge amount Ar, refund amount Ab, fund transfer amount At, complaint refund amount Af, and received balance Al.
[0181] The second part is the detailed information, expressed as D details, which covers the original detailed business data and system balance change detailed data during the stored-value card audit period, which is used to locate the specific calculation deviation point. The variable set in the detailed data includes the stored-value card number C, business time t, business type B, business status S, export serial number E, amount M and business date d.
[0182] Preferably, the input part comes from the following parts:
[0183] (1) Recharge, refund and transfer data generated by the stored-value card operation and management party;
[0184] (2) The original transaction data of traffic and expansion transactions of the road section owners and expansion application managers;
[0185] (3) Stored-value card settlement data of the provincial network center.
[0186] Step 4: Build a batch audit model to analyze the summary data to obtain a summary table, and generate a monthly audit report for stored-value cards based on the summary table, including:
[0187] Step 4.1, using the summary data as input, simulating the production system to calculate the system balance, determining the difference category and locating the problematic stored-value card by comparing the system balance;
[0188] Specifically, Figure 11-15 As shown in the figure, the original summary data of each stored-value card system is used to simulate the production system to calculate the system balance. The theoretical card face balance formula is "R theory = S beginning of period + (Ar - As + Ab - Al + Af)", and the system balance difference formula is "ΔR = R theory - S end of period". By comparing ΔR, the accuracy of the stored-value card system balance calculation is judged, and the problematic stored-value card is located. Further, by matching the summary data with the same category of data in the system balance change table, the difference category is preliminarily determined.
[0189] Step 4.2, the detailed data of the problem stored-value cards are compared by common identification, and the records of differences are generated to form a stored-value card batch audit difference summary table, and the stored-value card batch audit difference detailed table is matched according to the stored-value card batch audit difference summary table;
[0190] Specifically, Fig.16 As shown, for the stored-value cards with differences between the R theory and the S period end, the detailed data is compared item by item through the common identifiers E, d, and M to find the exact records that cause the differences.
[0191] Step 4.3, generate a monthly audit report based on the stored-value card batch audit difference summary table and the stored-value card batch audit difference detail table;
[0192] Specifically, the output results include the stored-value card batch audit difference summary table R difference summary and the stored-value card batch audit difference detail table R difference detail, which include some basic information of the input part, system original data information and system balance change table information, as well as the difference type T difference and difference amount ΔM obtained by analysis. Finally, a stored-value card monthly audit report is generated based on this information.
[0193] Step 4.4: Process the summary difference data output by the batch audit model, including:
[0194] The identified abnormal data is used to perform anomaly analysis based on the gradient boosting tree as follows:
[0195] ,
[0196] in, represents the input features, Indicates the exception category, represents the model parameters;
[0197] Specifically, the features Including abnormal amount , transaction type , customer level , and the environmental context in which the transaction occurs Output Includes exception categories such as "Data Error", "Suspected Fraud", "System Error", etc.
[0198] The decision-making algorithm defines the processing strategy according to the exception category, which is expressed as follows:
[0199] ,
[0200] in, Indicates the specific processing action to be taken, including "automatic correction", "request for review", "ignore", etc.; represents the decision tree parameters;
[0201] The continuous learning mechanism optimizes decision accuracy and processing efficiency, setting the optimization goals as:
[0202] ,
[0203] in, Indicates the actual processing result. represents the loss function, which is used to evaluate the deviation between the prediction strategy and the actual result. represents the i-th input feature;
[0204] Update models in real time and decision trees To reflect the latest processing effects and strategy performance. Implement online learning mechanisms, for example, use feedback loops to adjust the parameters of GBT and decision trees to ensure that the model continues to effectively reflect the latest data and trends.
[0205] Integrate into existing batch balance audit systems to ensure seamless data flow and feedback loops, monitor processing results in real time, and collect key performance indicators (KPIs) and user feedback.
[0206] For example, Fig.17 As shown in the figure, a cloud database is deployed in the Data Lake Governance Center on Huawei Cloud. The data synchronization tool DATAOS is used to synchronize the relevant data of the ORACLE database of the production system to the cloud database to complete the input part of the algorithm model. The code development of the calculation and analysis part of the algorithm model is carried out through scripts in the Data Lake Governance Center on Huawei Cloud. The abnormal balance analysis of the stored-value card is performed based on the data synchronized in step 1. After the abnormal balance analysis of the stored-value card is completed, the abnormal audit details and summary report of the stored-value card are formed, and a text analysis report is formed based on this.
[0207] In one or more embodiments, Fig.18 As shown, a system for checking the balance of an ETC stored-value card is disclosed, and the system comprises:
[0208] The single card audit model 101 is used to obtain the original data of the high-speed stored-value card and generate a stored-value card account; wherein the original data is the original detailed business data of all parties in the entire life cycle of the single stored-value card; the single card audit model is constructed to analyze the original data to obtain the abnormal detailed information of the single stored-value card; wherein the abnormal details include the stored-value card number, business time, abnormal type, abnormal amount and remarks, wherein the abnormal type and abnormal amount are summarized to form an analysis report of the single stored-value card, and the abnormal balance information is summarized and counted according to the abnormal type, including:
[0209] Step 2.1, initialize the stored-value card ledger data and add auxiliary analysis columns for analysis;
[0210] Step 2.2, further analyze the suspected abnormal location data of the stored-value card number, the starting sequence number and the ending sequence number to determine whether there is any abnormality in the data;
[0211] Step 2.3, cleaning the suspected abnormal location data to obtain abnormal location data;
[0212] Step 2.4: locate the abnormal flow according to the abnormal location data, and locate the abnormal features in the remarks;
[0213] Step 2.5: Output the abnormal details in the form of an analysis report of a single stored-value card and a summary of the abnormal balance information;
[0214] The batch audit model 102 is used to aggregate the original data to obtain aggregated data; wherein the aggregated data includes aggregated information and detailed information, the aggregated information includes the original aggregated business data during the stored-value card audit period and the aggregated data of the stored-value card system balance change record table and the beginning and ending system balance snapshots, and the detailed information includes the audit period, stored-value card number, settlement amount, recharge amount, refund amount, fund transfer amount, complaint refund amount and received balance; the batch audit model is constructed to analyze the aggregated data to obtain a aggregated table, and a monthly stored-value card audit report is generated according to the aggregated table, including:
[0215] Step 4.1, using the summary data as input, simulating the production system to calculate the system balance, determining the difference category and locating the problematic stored-value card by comparing the system balance;
[0216] Step 4.2, the detailed data of the problem stored-value cards are compared by common identification, and the records of differences are generated to form a stored-value card batch audit difference summary table, and the stored-value card batch audit difference detailed table is matched according to the stored-value card batch audit difference summary table;
[0217] Step 4.3: Generate a monthly audit report based on the stored-value card batch audit difference summary table and the stored-value card batch audit difference detail table.
[0218] The above disclosures are only some preferred embodiments of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiments and equivalent changes made according to the claims of the present invention are still within the scope of the invention.
Claims
1. A method for checking the balance of an ETC stored-value card, characterized in that: The method comprises the following steps: Step 1: Obtain the original data of the stored-value card and generate a stored-value card ledger; wherein the original data is the original detailed business data of all parties in the entire life cycle of the stored-value card; Step 2: Build a single card audit model to analyze the original data to obtain the abnormal details of the stored-value card; the abnormal details include the stored-value card number, business time, abnormal type, abnormal amount and remarks. The abnormal type and abnormal amount are summarized to form an analysis report of the stored-value card. At the same time, the abnormal balance information is summarized and counted according to the abnormal type, including: Step 2.1, initialize the stored-value card ledger data and add auxiliary analysis columns for analysis; Step 2.2, further analyze the suspected abnormal location data of the stored-value card number, the starting sequence number and the ending sequence number to determine whether there is any abnormality in the data; Step 2.3, cleaning the suspected abnormal location data to obtain abnormal location data; Step 2.4: locate the abnormal flow according to the abnormal location data, and locate the abnormal features in the remarks; Step 2.5: Output the abnormal details in the form of an analysis report of a single stored-value card and a summary of the abnormal balance information; Step 3: Aggregate the original data to obtain aggregated data; wherein the aggregated data includes aggregated information and detailed information, the aggregated information includes the original aggregated business data during the stored-value card audit period and the aggregated data of the stored-value card system balance change record table and the beginning and ending system balance snapshots, and the detailed information includes the audit period, stored-value card number, settlement amount, recharge amount, refund amount, fund transfer amount, complaint refund amount and received balance; Step 4: Build a batch audit model to analyze the summary data to obtain a summary table, and generate a monthly audit report for stored-value cards based on the summary table, including: Step 4.1, using the summary data as input, simulate the production system to calculate the theoretical system balance, locate the problematic stored-value card by comparing the difference between the theoretical system balance and the snapshot of the system balance at the end of the period, and determine the difference category by matching the summary data with the data of the same category in the system balance change table; Step 4.2, the detailed data of the problem stored-value cards are compared by common identification, and the records of differences are generated to form a stored-value card batch audit difference summary table, and the stored-value card batch audit difference detailed table is matched according to the stored-value card batch audit difference summary table; Step 4.3, generate a monthly audit report based on the stored-value card batch audit difference summary table and the stored-value card batch audit difference detail table; Wherein, the adding of auxiliary analysis column analysis includes: Calculate whether the first difference and the second difference are 0 within the transaction time T. 0 indicates that the transaction is normal, otherwise it indicates that the transaction is abnormal. The first difference is the difference between the sum of the card balance before the transaction and the transaction amount and the card balance after the transaction, and the second difference is the difference between the transaction amount and the next transaction amount. Initialize the pre-transaction balance and post-transaction balance of the transaction card; Calculate whether the first modified difference and the second modified difference are 0 within the transaction time T, 0 indicates that the transaction is normal, otherwise the transaction is abnormal; wherein the first modified difference is the difference between the sum of the pre-transaction balance and the transaction amount of the transaction card after initialization and the post-transaction balance of the transaction card after initialization, and the second modified difference is the difference between the post-transaction balance of the transaction card after initialization and the next transaction amount after initialization; When the first modified difference and the second modified difference are both 0, further check the suspected abnormal location data, where the suspected abnormal location data includes at least one of the following: A. If the difference after the first modification and the difference after the second modification show the characteristics of step-like opposite numbers, it is considered that the lane software has filled in the balance before the card transaction incorrectly in the uploaded flow; B. The Lane software incorrectly fills in the balance before / after the card transaction in the uploaded flow; C. The transaction date filled in incorrectly in the uploaded transaction records by the Lane Software; D. The Lane software filled in the wrong amount in the uploaded transaction data.
2. A method for checking the balance of an ETC stored-value card according to claim 1, characterized in that: The step 2 also includes the following steps: Step 2.6: Use reinforcement learning to process the abnormal details generated by the single card audit model, including: Take the feature vector of the abnormal details generated by the single card audit model as the state S, define the action set A and reward function ; Design objective function by combining deterministic policy gradient method and context-aware decision tree , which is expressed as follows: , in, Represents the policy parameters; Represents the policy function, which consists of parameters definition; According to the behavior strategy The resulting state distribution; represents the immediate reward obtained by taking action a in state s; is the value function of the next state; A coefficient that represents the balance between immediate rewards and long-term value; Designing a hybrid function The performance and stability of the strategy are optimized at the same time, as shown below: , in, Indicates the strategy Divergence ensures that the new strategy does not deviate too far from the old strategy and maintains the stability of the learning process; a represents the regularization parameter, which controls The effect of divergence on the total loss, represents the objective function; represents the old policy function, Represents the policy function; then use gradient ascent to update the policy parameters, as follows: , in, is the learning rate; Using a deterministic strategy Directly from the current state Generate actions.
3. A method for checking the balance of an ETC stored-value card according to claim 2, characterized in that: The state S includes at least one of the transaction amount, the difference in amount from the previous transaction, the time interval of the previous transaction, the transaction time of the day, the abnormal frequency in the account history, and the account level; the action set A includes at least one of automatically adjusting the account balance, sending an alarm to the user terminal for confirmation, and forwarding to manual review; wherein the reward function is designed as: , when the action correctly resolves the exception and minimizes user disruption; , when the action fails to resolve the exception or adds burden to the user; , when no further action is required.
4. The method for checking the balance of an ETC stored-value card according to claim 1, characterized in that: The ledger includes basic information, access information and card balance information; wherein, the basic information includes at least one of the stored-value card number C, business time t, business type B and business status S; the access information includes at least one of the access serial number, exit license plate, vehicle type, entry station, entry time Te, clearing time Tc and transaction location / exit station; the card balance information includes at least one of the receivable amount Ap, discount amount Ad, transaction amount Af, card balance before transaction Ab, card deduction amount At and card balance after transaction Aa.
5. The method for checking the balance of an ETC stored-value card according to claim 1, characterized in that: The data cleaning includes at least one of the following: a. For transactions where the difference between the first modification and the second modification presents a step-by-step opposite number, the transaction amount before and after the card transaction in the last line of abnormal transactions shall be used as the basis, and the data cleaning method of reverse chronological order shall be used for cleaning; b. For suspected abnormal transactions where the difference after the first modification and the difference after the second modification total 0, the balance before the card transaction in the first line of abnormal transaction shall be used as the basis and the data cleaning method of positive time sequence shall be used for cleaning; c. For suspected abnormal transactions with incorrect transaction time, if the entry time and transaction time T are close, the time will be calculated based on the clearing time - 1.
6. The method for checking the balance of an ETC stored-value card according to claim 1, characterized in that: The abnormal characteristics include at least one of the following: failure to deduct fees from the current transaction, failure to deduct fees from the upper and lower transactions, missing transactions, the first half of the recharge, and erroneous push of card limit adjustment data.
7. The method for checking the balance of an ETC stored-value card according to claim 1, characterized in that: The step 4 also includes, step 4.4, processing the summary difference data output by the batch audit model, including: The identified abnormal data is used to perform anomaly analysis based on the gradient boosting tree as follows: , Among them, F represents the input feature, C represents the abnormal category, represents the model parameters, represents the classification function; Based on the decision-making algorithm, the processing strategy is defined according to the exception category, as shown below: , Among them, A indicates the determination of specific processing actions; represents the decision tree parameters, represents the decision function; The continuous learning mechanism optimizes decision accuracy and processing efficiency, setting the optimization goals as: , in, Indicates the actual processing result. represents the loss function, which is used to evaluate the deviation between the prediction strategy and the actual result. represents the i-th input feature.
8. A system for checking the balance of an ETC stored-value card, characterized in that: The system comprises: A single card audit model is used to obtain the original data of a stored-value card and generate a stored-value card ledger; wherein the original data is the original detailed business data of all parties in the entire life cycle of a single stored-value card; a single card audit model is constructed to analyze the original data to obtain the abnormal detailed information of the single stored-value card; wherein the abnormal details include the stored-value card number, business time, abnormal type, abnormal amount and remarks, wherein the abnormal type and abnormal amount are summarized to form an analysis report of the single stored-value card, and the abnormal balance information is summarized and counted according to the abnormal type, including: Step 2.1, initialize the stored-value card ledger data and add auxiliary analysis columns for analysis; Step 2.2, further analyze the suspected abnormal location data of the stored-value card number, the starting sequence number and the ending sequence number to determine whether there is any abnormality in the data; Step 2.3, cleaning the suspected abnormal location data to obtain abnormal location data; Step 2.4: locate the abnormal flow according to the abnormal location data, and locate the abnormal features in the remarks; Step 2.5: Output the abnormal details in the form of an analysis report of a single stored-value card and a summary of the abnormal balance information; The batch audit model is used to aggregate the original data to obtain the aggregated data; wherein the aggregated data includes aggregated information and detailed information, the aggregated information includes the original aggregated business data during the stored-value card audit period and the aggregated data of the stored-value card system balance change record table and the snapshots of the system balance at the beginning and end of the period, and the detailed information includes the audit period, stored-value card number, settlement amount, recharge amount, refund amount, fund transfer amount, complaint refund amount and received balance; the batch audit model is constructed to analyze the aggregated data to obtain the aggregated table, and the monthly audit report of the stored-value card is generated according to the aggregated table, including: Step 4.1, using the summary data as input, simulate the production system to calculate the theoretical system balance, locate the problematic stored-value card by comparing the difference between the theoretical system balance and the snapshot of the system balance at the end of the period, and determine the difference category by matching the summary data with the data of the same category in the system balance change table; Step 4.2, the detailed data of the problem stored-value cards are compared by common identification, and the records of differences are generated to form a stored-value card batch audit difference summary table, and the stored-value card batch audit difference detailed table is matched according to the stored-value card batch audit difference summary table; Step 4.3, generate a monthly audit report based on the stored-value card batch audit difference summary table and the stored-value card batch audit difference detail table; Wherein, the adding of auxiliary analysis column analysis includes: Calculate whether the first difference and the second difference are 0 within the transaction time T. 0 indicates that the transaction is normal, otherwise it indicates that the transaction is abnormal. The first difference is the difference between the sum of the card balance before the transaction and the transaction amount and the card balance after the transaction, and the second difference is the difference between the transaction amount and the next transaction amount. Initialize the pre-transaction balance and post-transaction balance of the transaction card; Calculate whether the first modified difference and the second modified difference are 0 within the transaction time T, 0 indicates that the transaction is normal, otherwise the transaction is abnormal; wherein the first modified difference is the difference between the sum of the pre-transaction balance and the transaction amount of the transaction card after initialization and the post-transaction balance of the transaction card after initialization, and the second modified difference is the difference between the post-transaction balance of the transaction card after initialization and the next transaction amount after initialization; When the first modified difference and the second modified difference are both 0, further check the suspected abnormal location data, where the suspected abnormal location data includes at least one of the following: A. If the difference after the first modification and the difference after the second modification show the characteristics of step-like opposite numbers, it is considered that the lane software has filled in the balance before the card transaction incorrectly in the uploaded flow; B. The Lane software incorrectly fills in the balance before / after the card transaction in the uploaded flow; C. The transaction date filled in incorrectly in the uploaded transaction records by the Lane Software; D. The Lane software filled in the wrong amount in the uploaded transaction data.
