Automatic flow account checking method and electronic equipment

Abstract

The invention provides an automatic flow account checking method and electronic equipment, and the method comprises the steps: collecting transaction flow data, carrying out the evaluation of the transaction flow data, obtaining an evaluation result, carrying out the processing of the transaction flow data based on the evaluation result, and submitting the transaction flow data to a block chain for storage. And in response to the reconciliation triggering condition, extracting transaction flow data needing reconciliation from the block chain, and inputting the transaction flow data needing reconciliation into a reconciliation model obtained by pre-training for matching analysis processing to obtain a reconciliation result. If the account checking result shows that the consistency requirement is met, the account checking result is stored, and if the account checking result shows that the consistency requirement is not met, a difference report is generated, and an alarm mechanism is triggered. According to the scheme, mass data can be quickly analyzed and matched through the reconciliation model, and the time required for reconciliation is greatly shortened in combination with distributed storage and quick reading characteristics of the block chain.

Description

Technical Field

[0001] The present invention relates to the field of financial technology, and in particular to an automatic account reconciliation method and electronic equipment. Background Art

[0002] In many application scenarios, there is a need to check and match data. For example, in the financial field, transaction data needs to be reconciled. In current financial data processing, the methods commonly used to reconcile transaction data include manual reconciliation methods and automated software reconciliation methods.

[0003] Among them, in the manual reconciliation method, financial personnel manually export relevant flow data from various financial systems (such as the company's fund management system, business transaction system, etc.), and then use tools such as spreadsheets to check one by one. For example, financial personnel list the income flow and expenditure flow for a period of time separately, and then check whether each income has a corresponding expenditure explanation or whether each expenditure conforms to the business logic. Financial personnel use computer software to open the spreadsheet file that stores the flow data, and rely on manual visual recognition and brain judgment to perform reconciliation work, which is inefficient and prone to errors.

[0004] Therefore, the traditional manual reconciliation method has the defects of low efficiency, difficulty in ensuring accuracy, and lack of data security. Specifically, when the business volume of an enterprise is huge and the flow data is large, manual reconciliation requires a lot of time and manpower costs. After working for a long time, manual labor is prone to fatigue and negligence, resulting in reconciliation errors, and some important difference data may be missed or different data may be matched incorrectly. Flow data in paper and electronic form is prone to data leakage and loss during transmission and storage.

[0005] In the automated software reconciliation method, there are some automated financial software that matches the transaction flow by setting some fixed rules, such as equal amounts, dates within a certain range, etc. For example, in e-commerce companies, the software automatically checks whether the order amount is consistent with the amount received by the payment platform. Data enters the system from different data sources (such as business databases, third-party payment interfaces, etc.), and the rule engine processes and matches the data according to the set rules (such as amount thresholds, etc.), and outputs and displays the successful and unsuccessful matching results separately.

[0006] Therefore, the automated software reconciliation method has defects such as rule limitations, lack of flexibility, and data trust issues. Specifically, it can only handle scenarios that comply with predefined rules, and it is difficult to accurately match complex business logic and special situations (such as installment payments across months, the existence of handling fees, etc.). When the business model changes or new reconciliation requirements arise, it is necessary to modify the rules, adjust the software parameters, or even redevelop some modules, which is costly and time-consuming. The data accuracy of different data sources is difficult to verify. When there is a reconciliation inconsistency, it is difficult to determine which data source has a problem. There is no reliable mechanism to ensure the immutability and traceability of the data. Summary of the invention

[0007] The purpose of the embodiments of the present invention is to provide an automatic transaction flow reconciliation method and electronic device, so as to quickly and accurately reconcile transaction flow data.

[0008] In a first aspect, the present invention provides an automatic account reconciliation method, the method comprising:

[0009] Collecting transaction flow data, evaluating the transaction flow data to obtain an evaluation result, processing the transaction flow data based on the evaluation result, and submitting the transaction flow data to the blockchain for storage;

[0010] In response to the reconciliation trigger condition, the transaction flow data that needs to be reconciled is extracted from the blockchain, and the transaction flow data that needs to be reconciled is input into the pre-trained reconciliation model for matching analysis and processing to obtain the reconciliation result;

[0011] If the reconciliation result shows that the transaction flow data to be reconciled meets the consistency requirement, the reconciliation result is stored in a specific location in the blockchain;

[0012] If the reconciliation result shows that the transaction flow data to be reconciled does not meet the consistency requirements, a difference report is generated and an alarm mechanism is triggered.

[0013] In an optional implementation manner, the step of evaluating the transaction flow data to obtain an evaluation result, and processing the transaction flow data based on the evaluation result includes:

[0014] Evaluate the transaction flow data from multiple dimensions to obtain evaluation results of each dimension, wherein the multiple dimensions include completeness dimension, accuracy dimension, consistency dimension and timeliness dimension;

[0015] Calculate the total evaluation value according to the evaluation results of each dimension, and filter out abnormal data in the transaction flow data according to the total evaluation value;

[0016] The abnormal data is automatically corrected or marked.

[0017] In an optional implementation, the step of collecting transaction flow data includes:

[0018] The collection tasks are allocated through the master control node, wherein the collection tasks are allocated according to any one of data source division, time range division and data volume division;

[0019] The transaction flow data is collected by multiple working nodes according to the allocation of collection tasks, and the master control node records the execution status of each working node;

[0020] The transaction flow data collected by each working node is stored through the data storage node.

[0021] In an optional implementation, the step of submitting the transaction flow data to the blockchain for storage includes:

[0022] Grouping the transaction flow data in a set manner and generating a data fingerprint for each group of transaction flow data;

[0023] Store each set of transaction flow data in the block body of each block of the blockchain in order, generate a hash value based on the block header and block body of the previous block, and store the hash value of the previous block in the block header of the next block;

[0024] The full node in the blockchain is used to store the information of each block, and the light node in the blockchain is used to save the block header information of each block. The full node can be used to find the block where the required reconciliation transaction flow data is located.

[0025] In an optional implementation, before the step of responding to the reconciliation trigger condition, the method further includes a step of verifying the transaction flow data in the blockchain, the step including:

[0026] Requesting the Merkle proof of the transaction flow data to be verified from the full node through the light node;

[0027] Find the block where the transaction flow data is located through the full node, and construct a path from the transaction flow data to the Merkle root of the block, wherein the path from the transaction flow data to the Merkle root of the block is a Merkle proof;

[0028] After receiving the Merkle proof, the light node uses the hash value in the Merkle proof to recalculate the Merkle root, and compares the calculated Merkle root with the Merkle root in the saved block header of the block. If the two are consistent, the verification is determined to be successful.

[0029] In an optional embodiment, before the step of inputting the transaction flow data to be reconciled into the pre-trained reconciliation model for matching analysis and processing, the method further includes a step of pre-processing the transaction flow data to be reconciled, which step includes at least one of the following:

[0030] Standardize the numerical features of transaction flow data that need to be reconciled;

[0031] Performing feature coding processing on the transaction flow data;

[0032] Converting the timestamp of the transaction flow data into a periodic time feature;

[0033] An interaction feature is created based on the transaction flow data, wherein the interaction feature includes an amount difference and a time difference.

[0034] In an optional implementation, the reconciliation model includes a first sub-model and a second sub-model;

[0035] The step of inputting the transaction flow data to be reconciled into the pre-trained reconciliation model for matching analysis and processing includes:

[0036] Input the transaction flow data that needs to be reconciled into the first sub-model and the second sub-model that have been trained in advance respectively;

[0037] Performing matching analysis and processing through the first sub-model to obtain a first reconciliation result, and performing matching analysis and processing through the second sub-model to obtain a second reconciliation result;

[0038] The first reconciliation result and the second reconciliation result are weighted averaged to obtain the final reconciliation result.

[0039] In an optional implementation, after the step of acquiring transaction flow data, the method further includes a step of identifying noise in the transaction flow data, which step includes at least one of the following:

[0040] Using statistical methods to identify transaction flow data with abnormal values ​​in the transaction flow data;

[0041] Transaction flow data that does not satisfy predefined business rules is identified in the transaction flow data.

[0042] In an optional implementation, after the step of acquiring transaction flow data, the method further includes a step of identifying abnormal values ​​in the transaction flow data, which step includes at least one of the following:

[0043] Detecting outliers from the transaction flow data using statistical methods;

[0044] The transaction flow data with abnormal fluctuations is detected by using a moving average detection method for the time series in the transaction flow data.

[0045] In a second aspect, the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps of the method described in any one of the aforementioned implementation modes are implemented.

[0046] The present invention provides an automatic transaction flow reconciliation method and electronic device, which collects transaction flow data, evaluates the transaction flow data to obtain an evaluation result, processes the transaction flow data based on the evaluation result, and then submits the transaction flow data to the blockchain for storage. In response to the reconciliation trigger condition, the transaction flow data that needs to be reconciled is extracted from the blockchain, and the transaction flow data that needs to be reconciled is input into a pre-trained reconciliation model for matching analysis and processing to obtain a reconciliation result. If the reconciliation result shows that the consistency requirement is met, the reconciliation result is stored. If the reconciliation result shows that the consistency requirement is not met, a difference report is generated and an alarm mechanism is triggered. This solution can quickly analyze and match massive data through the reconciliation model, and combined with the distributed storage and fast reading characteristics of the blockchain, the time required for reconciliation is greatly shortened. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments of the present invention are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 A flow chart of an automatic account reconciliation method provided by an embodiment of the present invention;

[0049] Figure 2 for Figure 1 A flowchart of the sub-steps included in S11;

[0050] Figure 3 for Figure 1 Another flow chart of the sub-steps included in S11;

[0051] Figure 4 for Figure 1 Another flow chart of the sub-steps included in S11;

[0052] Figure 5 A flowchart of a verification method in the automatic account reconciliation method provided by an embodiment of the present invention;

[0053] Figure 6 for Figure 1 A flowchart of the sub-steps included in S12;

[0054] Figure 7 A functional module block diagram of an automatic account reconciliation device provided by an embodiment of the present invention;

[0055] Figure 8 This is a structural block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present invention will be described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0057] See also Figure 1 , is a flow chart of an automatic account reconciliation method provided by an embodiment of the present invention, the automatic account reconciliation method can be implemented by an automatic account reconciliation device, the automatic account reconciliation device can be implemented by software and / or hardware, and can be configured in an electronic device, the electronic device can be a computer device, a server, etc. The detailed steps of the automatic account reconciliation method are described as follows.

[0058] S11, collecting transaction flow data, evaluating the transaction flow data to obtain an evaluation result, processing the transaction flow data based on the evaluation result, and submitting the transaction flow data to the blockchain for storage.

[0059] S12, in response to the reconciliation trigger condition, extracting the transaction flow data that needs to be reconciled from the blockchain, inputting the transaction flow data that needs to be reconciled into the pre-trained reconciliation model for matching analysis and processing, and obtaining the reconciliation result.

[0060] S13: If the reconciliation result indicates that the transaction flow data to be reconciled meets the consistency requirement, the reconciliation result is stored in a specific location in the blockchain.

[0061] S14: If the reconciliation result shows that the transaction flow data to be reconciled does not meet the consistency requirement, a difference report is generated and an alarm mechanism is triggered.

[0062] In this embodiment, transaction flow data can be collected from multiple data sources, including data sources such as enterprise financial systems, bank transaction interfaces, e-commerce platform payment economic records, etc. Transaction flow data can be collected through API interfaces or data import tools of corresponding data sources.

[0063] In this embodiment, the collection of transaction flow data can be achieved based on a distributed data collection architecture, supporting high-concurrency data capture.

[0064] The collected transaction flow data is evaluated mainly from the perspectives of completeness, accuracy, consistency and timeliness of the transaction flow data. The transaction flow data is processed based on the evaluation results. For example, the transaction flow data that does not meet the relevant requirements can be corrected, or the data can be marked and provided to the staff for verification and adjustment.

[0065] On this basis, the transaction flow data is submitted to the blockchain for preservation. The blockchain is a chain structure composed of multiple blocks. Using the blockchain to save transaction flow data can ensure that the data is not tampered with, and because the blockchain has a backup mechanism, data loss can be avoided.

[0066] When reconciliation is required, in response to a reconciliation trigger condition, the reconciliation trigger condition can be a timed trigger condition or an event trigger condition. For example, it can be set to trigger the reconciliation task at a fixed interval (such as every morning), or it can be set to trigger the reconciliation task when new transaction flow data is stored in the blockchain and reaches a certain amount of data, or it can be triggered when an external reconciliation instruction is received.

[0067] After the reconciliation task is triggered, the transaction flow data that needs to be reconciled is extracted from the blockchain. The transaction flow data that needs to be reconciled can be one or more groups of transaction flow data, and each group of transaction flow data includes two transaction flow data.

[0068] The transaction flow data that needs to be reconciled is input into a pre-trained reconciliation model, which is a pre-trained model used to match two input transaction flow data to output a reconciliation result of the matching degree between the two.

[0069] The purpose of reconciling transaction flow data is to determine whether the two parties are consistent. Therefore, if the reconciliation result shows that the transaction flow data to be reconciled meets the consistency requirement, it means that the transaction flow data for reconciliation is consistent. Therefore, the reconciliation result can be saved in a specific location, such as the blockchain ledger of the blockchain, for subsequent query.

[0070] If the reconciliation result shows that the transaction flow data that needs to be reconciled does not meet the consistency requirements, it means that the transaction flow data for reconciliation is inconsistent. This situation indicates that there may be anomalies in the transaction flow data, so a difference report can be generated, which can include the specific differences between the transaction flow data that do not meet the consistency requirements and trigger an alarm mechanism, so that the staff can be prompted to check the transaction flow data that has anomalies and does not meet the consistency requirements.

[0071] The automatic banknote reconciliation method provided in this embodiment can quickly analyze and match massive data through the reconciliation model, and combined with the distributed storage and fast reading characteristics of the blockchain, it greatly shortens the time required for reconciliation.

[0072] In this embodiment, in the above-mentioned step of collecting transaction flow data, a distributed architecture is used to implement the collection. The distributed architecture is composed of multiple computers or servers, and the multiple computers or servers include a master control node, a working node, and a data storage node. Figure 2 , the steps of collecting transaction flow data can be achieved in the following ways:

[0073] S111, allocating the collection tasks through the master control node, wherein the collection tasks are allocated according to any one of data source division, time range division and data volume division.

[0074] S112, collecting transaction flow data through multiple working nodes according to the allocation of collection tasks, and the main control node records the execution status of each working node.

[0075] S113, storing the transaction flow data collected by each working node through the data storage node.

[0076] In this embodiment, the master control node in the distributed architecture is responsible for task allocation and coordination, each working node is used to perform actual data collection tasks, and the data storage node is used to store the collected transaction flow data. The master control node can be used to divide the large collection task into multiple small tasks, and then assign these small tasks to different working nodes. Each working node performs its own task at the same time, and finally the master control node collects and organizes the execution status of all working nodes.

[0077] When the master control node allocates tasks, it can be divided according to data source, time range or data volume. For example, different work nodes can be responsible for different data sources, such as some work nodes can be responsible for ERP system, and some work nodes can be responsible for bank transaction data system. Or, for a large amount of historical transaction flow data, the collection of transaction flow data in different time periods can be allocated to different work nodes according to the division of time periods. Or it can be evenly distributed to each work node according to the amount of data to ensure the balanced distribution of collection tasks.

[0078] In this embodiment, in the process of collecting transaction flow data based on a distributed architecture, task allocation can be dynamically adjusted to ensure balanced workload of each working node. In addition, when a working node failure is detected, tasks can be reallocated to ensure the continuity of data collection. In order to ensure data consistency, a distributed lock or version control mechanism can be used to avoid duplicate collection or data conflicts.

[0079] For the collected transaction flow data, noise in the transaction flow data can be identified. Noise refers to random errors or irrelevant information in the data. Specifically, noise in the transaction flow data can be identified by at least one of the following methods:

[0080] Using statistical methods to identify transaction flow data with abnormal values ​​in the transaction flow data;

[0081] Transaction flow data that does not satisfy predefined business rules is identified in the transaction flow data.

[0082] In this embodiment, a statistical method, such as a Z-score method, can be used to identify abnormal values ​​in numerical data in the transaction flow data, and the abnormal values ​​are used as noise. In addition, predefined business rules are applied to identify transaction flow data that do not meet the business rules as noise. For example, the predefined business rules include that the transaction amount should not be a negative value.

[0083] The identified noise can be deleted, corrected, or marked as pending, depending on the situation.

[0084] In addition, in this embodiment, outliers in the collected transaction flow data can also be identified. Outliers refer to data points that significantly deviate from the normal pattern. Specifically, outliers in the transaction flow data can be identified by at least one of the following methods:

[0085] Detect outliers from the transaction flow data using statistical methods.

[0086] The transaction flow data with abnormal fluctuations is detected by using a moving average detection method for the time series in the transaction flow data.

[0087] The statistical method used here can be a box plot (IQR method) to detect outliers in the transaction flow data as abnormal values. In addition, for time series data, a moving average can be used to detect abnormal values ​​therein.

[0088] The detected outliers can be marked for further analysis.

[0089] In addition, in this embodiment, the transaction flow data needs to be converted to achieve a unified format, including date standardization, amount standardization, text encoding, field naming, etc.

[0090] In date standardization, you can convert it to ISO 8601 format (YYYY-MM-DDTHH:mm:ss.sssZ). In amount standardization, you can unify the currency unit and decimal places. In text encoding, ensure that all texts use a unified character encoding (such as UTF-8). In field naming, you can use consistent naming rules (such as camel case naming).

[0091] After completing the above-mentioned identification and processing of noise and abnormal values ​​in the transaction flow data, as well as the format standardization conversion, the transaction flow data can be evaluated to obtain an evaluation result, and the transaction flow data can be processed based on the evaluation result. Figure 3 , this step can be achieved by:

[0092] S114, evaluating the transaction flow data from multiple dimensions to obtain evaluation results of each dimension, wherein the multiple dimensions include a completeness dimension, an accuracy dimension, a consistency dimension, and a timeliness dimension.

[0093] S115, calculating a total evaluation value according to the evaluation results of each dimension, and filtering out abnormal data in the transaction flow data according to the total evaluation value.

[0094] S116, automatically correcting the abnormal data or marking the abnormal data.

[0095] In this embodiment, when evaluating transaction flow data, it can be evaluated from the dimensions of completeness, accuracy, consistency, and timeliness.

[0096] Among them, the integrity dimension mainly detects whether the necessary fields in the transaction flow data exist and whether they have values. For example, the necessary fields can be fields such as transaction ID, amount, date, etc. In addition, the proportion of non-empty fields in each transaction record can also be detected. The higher the proportion of non-empty fields, the higher the integrity.

[0097] The accuracy dimension mainly detects whether the values ​​of transaction flow data are within a reasonable range and whether they comply with the prescribed format. For example, it detects whether the transaction amount is within a reasonable range, such as negative numbers or abnormally large values ​​should not appear. Another example is whether the date format is correct and whether the account number complies with the prescribed format. In addition, it can detect whether the debits and credits are balanced, and whether the information of the payee and payee matches.

[0098] The consistency dimension mainly detects whether the same transaction data across different data sources are consistent. For example, it compares whether the same transaction data from different systems, such as ERP (Enterprise Resource Planning) system and bank statement are consistent. In addition, the consistency ratio can also be calculated, which can be obtained by dividing the number of consistent records by the total number of records.

[0099] The timeliness dimension is mainly to detect whether the data is up to date and within the expected time range. You can check the timestamp of the transaction flow data to ensure that the transaction flow data is within the expected time range. In addition, you can calculate the difference between the data collection time and the actual transaction time to get the data delay time.

[0100] The transaction flow data obtains corresponding evaluation results in each dimension, and the evaluation results of each dimension are accumulated according to the set weights to obtain the evaluation summary. For example, the total evaluation value can be calculated in the following way:

[0101] Total evaluation value = completeness score * 30% + accuracy score * 40% + consistency score * 20% + timeliness score * 10%

[0102] Based on the obtained total evaluation value, abnormal data in the transaction flow data is screened out. For example, a quality threshold can be set in advance, and the transaction flow data with a total evaluation value less than the quality threshold is regarded as abnormal data.

[0103] For the identified abnormal data, the processing method can be that the system automatically corrects it, or marks it for staff to check and process.

[0104] For example, the system can automatically correct transaction data with obvious errors, such as inconsistent date formats. Abnormal data that requires manual intervention, such as consistency issues, can be marked for subsequent processing. A task list can be generated for marked abnormal data to avoid missing abnormal data.

[0105] In addition, in this embodiment, all evaluation results and processing operations can be recorded for subsequent auditing and system optimization.

[0106] In this embodiment, the evaluation process can also be continuously optimized, including, for example, regularly analyzing the patterns and trends of data quality issues, adjusting the evaluation rules and weights based on the analysis results, and taking preventive measures at the source for frequently occurring problems.

[0107] After the transaction flow data is identified and processed as above, the transaction flow data can be submitted to the blockchain for storage. Figure 4 In this embodiment, this step can be implemented by:

[0108] S117, grouping the transaction flow data according to a set method, and generating a data fingerprint for each group of transaction flow data.

[0109] S118, storing each group of transaction flow data in the block body of each block of the blockchain in sequence, generating a hash value based on the block header and block body in the previous block, and storing the hash value of the previous block in the block header of the next block.

[0110] S119, using the full node in the blockchain to store the information of each of the blocks, and using the light node in the blockchain to save the block header information of each of the blocks, the full node can be used to find the block where the required reconciliation transaction flow data is located.

[0111] In this embodiment, the processed transaction flow data is stored in the blockchain in the form of blocks. Each block contains the transaction flow data within a certain period of time and is connected to the previous block through a hash algorithm to form an unalterable chain structure.

[0112] In this embodiment, Hyperledger Fabric (an open source enterprise-level blockchain technology) can be used as the blockchain platform, and the PBFT (Practical Byzantine Fault Tolerance) consensus algorithm is used to meet high consistency requirements.

[0113] The transaction flow data is grouped in a set manner, for example, by time interval, such as hourly or daily, or by data volume, such as every 1,000 records. Each group of transaction flow data is hashed to generate a unique data fingerprint.

[0114] Each set of transaction flow data is stored in each block in the blockchain in order. Each block includes a block header and a block body. The block header in each block contains the hash value of the previous block, timestamp, block height (serial number), and Merkle root (the root hash of the Merkle tree constructed by all transaction hashes of the block). The block body in each block includes all transaction flow data records within the time period and the hash value of each record.

[0115] When all blocks are linked into a chain structure, the hash value of the new block is calculated (based on its block header and block body), and then the hash value of the new block is stored in the block header of the next block to form a chain structure.

[0116] When a new block is added to the blockchain, it needs to be verified by consensus. In this embodiment, the PBFT consensus algorithm can be used to implement consensus verification. After the new block is verified by consensus, it is added to the blockchain.

[0117] In addition, the transaction flow data added to the block can also be encrypted, for example, using the AES-256 algorithm to ensure the confidentiality of the data on the blockchain.

[0118] In addition, the blockchain also includes full nodes and light nodes. Full nodes are used to store information of all blocks. They are complete blockchain databases that contain all information from the genesis block to the latest block. Light nodes are used to save the information of the block headers of each block.

[0119] Among them, the full node can be understood as a full node ledger, in which the information stored can be expressed as: block 1 + block 2 + ... + block n, and the information contained in each block can be expressed as: block header + block body (transaction flow data).

[0120] Through the full node, you can access transaction flow data at any point in time, and quickly locate the transaction flow data you need to query or extract according to the block height or timestamp.

[0121] On the basis of storing the collected transaction flow data in the blockchain, when reconciliation is required, the transaction flow data that needs to be reconciled can be extracted from the blockchain. In this embodiment, in order to ensure the integrity and non-tamperability of all transaction flow data, before actually performing the reconciliation, the transaction flow data in the blockchain can also be verified. This step can be achieved in the following way. Please refer to Figure 5 :

[0122] S21, requesting the Merkle proof of the transaction flow data that needs to be verified from the full node through the light node.

[0123] S22, searching the block where the transaction flow data is located through the full node, and constructing a path from the transaction flow data to the Merkle root of the block, wherein the path from the transaction flow data to the Merkle root of the block is a Merkle proof.

[0124] S23, after receiving the Merkle certificate, the light node uses the hash value in the Merkle certificate to recalculate the Merkle root, and compares the calculated Merkle root with the Merkle root in the saved block header of the block. If the two are consistent, it is determined that the verification is successful.

[0125] In this embodiment, the light node only saves the block header information of each block, and can periodically obtain the latest block header information from the full node. When a transaction flow data needs to be verified, the light node can request the Merkle proof of the transaction flow data from the full node.

[0126] The full node can find the block where the transaction flow data to be verified is located, and construct a path from the transaction flow data to the block Merkle root, that is, the Merkle proof. The full node can feedback the constructed Merkle proof to the light node. After receiving the Merkle proof, the light node uses the hash value in the received Merkle proof to recalculate the Merkle root, and compares the calculated Merkle root with the Merkle root in the block header of the stored corresponding block. If the calculated Merkle root is consistent with the Merkle root in the block header, it proves that the transaction flow data does exist in the block and has not been tampered with. If the two are inconsistent, it means that the transaction flow data may have been tampered with or does not exist in the block.

[0127] In addition, in this embodiment, the light node can also check the link relationship of the block header to ensure the integrity of the blockchain.

[0128] After passing the above verification, when the reconciliation trigger condition is triggered, the reconciliation trigger condition can be responded to, the transaction flow data that needs to be reconciled can be extracted from the blockchain, and the transaction flow data that needs to be reconciled can be input into the pre-trained reconciliation model for matching analysis and processing to obtain the reconciliation result.

[0129] In this embodiment, the process of pre-training the reconciliation model is first described below.

[0130] This embodiment can use machine learning and deep learning algorithms to train historical transaction flow data to build a reconciliation model. The trained reconciliation model can automatically identify the characteristics and patterns of data in different data sources and perform accurate matching and comparison. For complex transaction scenarios and abnormal situations, intelligent analysis and judgment can be performed to improve the accuracy of reconciliation.

[0131] In this embodiment, a reconciliation model can be built based on the TensorFlow framework, and random forest, neural network algorithms, etc. can be used. The training data can be updated regularly to adapt to business changes, and anomaly detection mechanisms can be introduced to timely discover potential reconciliation risks.

[0132] In this embodiment, the reconciliation model can be constructed based on any one of the random forest model and the neural network model, or a combination of the two.

[0133] Among them, the random forest model is an integrated learning method composed of multiple decision trees. It is suitable for processing various types of features and is insensitive to outliers.

[0134] First, the parameters of the random forest model can be set, including the number of decision trees (such as 100), the maximum depth (unlimited, so that each decision tree can grow fully), the minimum number of leaf node samples (for example, 5), feature selection (for example, sqrt(n_features)), sample sampling (for example, bootstrap sampling can be used), and other settings.

[0135] In this embodiment, a multi-layer perceptron can be used as a neural network model, which can learn complex nonlinear relationships.

[0136] The neural network model is hierarchically structured, including an input layer, a hidden layer 1, a hidden layer 2, and an output layer. The input layer can be configured with nodes based on the number of features, for example, 64 nodes can be configured. Hidden layer 1 can include 128 nodes, and a ReLU activation method can be used. Hidden layer 2 can include 64 nodes, and a ReLU activation method can be used. The output layer can include 1 node, and a Sigmoid activation method can be used.

[0137] When constructing training samples, you can load historical transaction flow data and perform feature engineering on the collected historical transaction flow data, where feature engineering includes standardization of numerical features, encoding of type features, extraction of time features, and construction of interactive features. Standard Scaler method can be used to standardize numerical features. In the encoding of category features, Label Encoding method can be used for random forest models, and One-Hot encoding method can be used for neural network models. In the extraction of time features, timestamps can be converted into periodic features, such as time of day, day of the week, etc. In the creation of interactive features, for example, amount differences, time differences, etc. can be constructed.

[0138] The historical transaction flow data processed as above is divided into a training set and a validation set, for example, the ratio can be 8:2.

[0139] When training the random forest model using the training set, the model performance can be evaluated using cross-validation evaluation, and the model hyperparameters can be optimized through grid search.

[0140] When training the neural network model using the training set, the binary cross entropy loss function and the Adam optimizer can be used for training, and the early stopping method can be used to prevent the problem of overfitting in training. The learning rate decay can be used to improve the convergence of the model.

[0141] Based on the training, the validation set can be used to evaluate the performance of the trained model, which can be evaluated by calculating indicators such as accuracy, precision, recall and F1 score.

[0142] Through the above method, when it is detected that the performance of the trained model meets the requirements, the model can be fixed for use in the analysis of transaction flow data that needs to be reconciled in the subsequent practical application stage.

[0143] In addition, the trained model can be updated and maintained, such as regular retraining, using the newly accumulated transaction flow data every month to retrain the model. Continuously monitor the performance of the reconciliation model in actual applications, implement reconciliation model deviation detection, and trigger alarms when performance degrades.

[0144] In the actual application stage, after obtaining the transaction flow data that needs to be reconciled, the transaction flow data can be input into the reconciliation model for processing.

[0145] In order to improve the quality of data input into the reconciliation model, in this embodiment, the transaction flow data to be reconciled may be preprocessed first, and the preprocessing step may include at least one of the following:

[0146] Standardize the numerical features of transaction flow data that need to be reconciled;

[0147] Performing feature coding processing on the transaction flow data;

[0148] Converting the timestamp of the transaction flow data into a periodic time feature;

[0149] An interaction feature is created based on the transaction flow data, wherein the interaction feature includes an amount difference and a time difference.

[0150] In this embodiment, similar to the above-mentioned method of performing feature processing on training samples in the model training stage, the transaction flow data that needs to be reconciled can be pre-processed from multiple aspects.

[0151] When performing numerical feature standardization, the transaction amount and account balance can be standardized.

[0152] When performing feature encoding, categorical features can be encoded, including transaction type, account type, transaction status, etc. Word vectorization encoding can be used for text features, such as transaction description, transaction party information, etc.

[0153] When processing timestamps, for example, the timestamps can be converted into periodic time features, such as the time point of the day, the day of the week, and so on.

[0154] When creating interactive features, you can create features such as the amount difference, time difference, and historical matching rate of two transaction flow data.

[0155] After the above preprocessing, assuming that the transaction flow data to be reconciled includes transaction A and transaction B, the features of transaction A and transaction B can be obtained as follows:

[0156] input_data=[

[0157] [1000.00,1693649400,2,1,0.98,0,24.5,...],#Characteristics of transaction A

[0158] [1000.00,1693735800,2,3,0.97,1,24.5,...],#Characteristics of transaction B ]

[0160] In this embodiment, the transaction flow data to be reconciled is input into the reconciliation model, and the reconciliation model performs matching analysis and processing, and finally outputs the matching probability of the two. The matching probability is a value between 0 and 1, indicating the possibility of matching of the two transaction flow data. For example, 0.95 indicates that there is a 95% probability that the two transaction flow data are matched. In addition, a confidence score can also be output, which indicates the degree of confidence of the reconciliation model in the reconciliation result it outputs.

[0161] The reconciliation results finally output by the reconciliation model can be shown as follows:

[0162] output={

[0163] "match_probability":0.95,

[0164] "match_class":1,

[0165] "confidence_score":0.92,

[0166] "potential_discrepancy":None

[0167] }

[0168] As can be seen from the above, the reconciliation model can be composed of a random forest model and a neural network model. In this embodiment, they are named as the first sub-model and the second sub-model respectively.

[0169] In the above step of inputting the transaction flow data to be reconciled into the pre-trained reconciliation model for matching analysis and processing, this can be achieved in the following ways. Please refer to Figure 6 :

[0170] S121, inputting the transaction flow data that needs to be reconciled into the first sub-model and the second sub-model that are trained in advance respectively.

[0171] S122, performing matching analysis and processing through the first sub-model to obtain a first reconciliation result, and performing matching analysis and processing through the second sub-model to obtain a second reconciliation result.

[0172] S123: Perform weighted average processing on the first reconciliation result and the second reconciliation result to obtain a final reconciliation result.

[0173] In this embodiment, the first sub-model and the second sub-model can be used to match and analyze the transaction flow data that needs to be reconciled, respectively, to obtain the first reconciliation result and the second reconciliation result. The first reconciliation result and the second reconciliation result are weighted averaged according to the set weights to obtain the final reconciliation result.

[0174] In this embodiment, after judging whether the transaction flow data that needs to be reconciled meets the consistency requirements based on the reconciliation results, the final judgment results can be sorted and summarized and output to the user in the form of visual reports, data files, etc., to clearly display the reconciled successful and unmatched transactions.

[0175] In this embodiment, HTML5 and JavaScript technology can be used to build the front-end interface, supporting multiple chart types (such as bar charts, line charts, pie charts, etc.) to display the reconciliation results. Provide data export function, support CSV, Excel and other formats.

[0176] In this way, the front-end interface interacts with the back-end through HTTP / HTTPS in the form of a web application, displays information such as the reconciliation task list, progress, results, and provides data visualization charts.

[0177] In addition, an interface for querying reconciliation results can be provided to external systems. The interface uses RESTful style or RPC, etc. to perform access control and data encryption transmission.

[0178] The automatic reconciliation method provided in this embodiment is compared with the existing manual reconciliation or partially automated software reconciliation, which often takes a lot of time. The use of AI technology in this solution can quickly analyze and match massive data, and combined with the distributed storage and fast reading characteristics of blockchain, it greatly shortens the time required for reconciliation. For example, for a company with thousands of transactions, traditional reconciliation may take several hours, while this system can be completed in seconds or even less.

[0179] In addition, compared with the existing technology, manual reconciliation is prone to errors due to fatigue or negligence, and some automated software reconciliation is limited by fixed rules and has limited ability to handle complex situations. In this solution, the AI ​​model is trained with a large amount of historical data and can identify complex transaction patterns and abnormal situations. The tamper-proof nature of the blockchain ensures the originality and authenticity of the data, thereby improving the accuracy of reconciliation. For example, it can accurately identify subtle differences caused by exchange rate fluctuations, fee calculations, etc.

[0180] In addition, compared with the existing technology, data is usually stored in a centralized database, which is vulnerable to attack and tampering, and there may be risks in the data transmission process. In this solution, the encryption and distributed storage mechanism of the blockchain makes it almost impossible to tamper with the data, and the data is synchronously backed up on multiple nodes. Even if some nodes fail, the integrity of the data is not affected. At the same time, advanced encryption algorithms are used in the data transmission process to effectively prevent data leakage.

[0181] Furthermore, compared with the existing technology, when business rules or data sources change, the software may need to be modified or even redeveloped on a large scale. In this solution, the AI ​​model can adapt to new business models and rules by continuously updating training data. The blockchain architecture also facilitates the addition of new nodes and data sources, and the system can easily cope with the development and changes of enterprise business.

[0182] Furthermore, compared with the existing technology, manual reconciliation requires a lot of manpower investment, and the maintenance and upgrade of automated software also requires a certain amount of cost. In this solution, efficient and accurate reconciliation reduces labor costs, while the scalability and adaptability of the system reduce long-term maintenance and upgrade costs. For example, enterprises do not need to frequently recruit and train a large number of reconciliation personnel, nor do they need to pay high fees for frequent software changes.

[0183] Furthermore, compared with the existing technology, the reconciliation results are often simply a match or not, which is difficult to provide in-depth analysis and insights. In this solution, through in-depth mining and analysis of reconciliation data, it can provide decision support for corporate management in terms of financial process optimization, risk control, etc. For example, it can discover the types of reconciliation differences that frequently occur, so as to improve the business process in a targeted manner.

[0184] In summary, by combining blockchain and AI technology, the present invention is significantly superior to existing technologies in terms of reconciliation efficiency, accuracy, security, flexibility, cost and decision support, bringing more reliable, efficient and intelligent solutions to the financial management of enterprises.

[0185] Based on the same inventive concept, please refer to Figure 7 , shows a functional module schematic diagram of the automatic account reconciliation device provided by an embodiment of the present invention. This embodiment can divide the functional modules of the automatic account reconciliation device according to the above method embodiment. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware or in the form of software functional modules.

[0186] It should be noted that the division of modules in the embodiments of the present invention is schematic and is only a logical function division. There may be other division methods in actual implementation. For example, in the case of dividing each functional module according to each function, Figure 7 The automatic account reconciliation device shown is only a schematic diagram of the device. The automatic account reconciliation device may include a collection module and a processing module. The functions of each functional module of the automatic account reconciliation device are described in detail below.

[0187] An evaluation module, used to collect transaction flow data, evaluate the transaction flow data to obtain an evaluation result, process the transaction flow data based on the evaluation result, and submit the transaction flow data to the blockchain for storage;

[0188] A processing module, which is used to respond to a reconciliation trigger condition, extract transaction flow data that needs to be reconciled from the blockchain, input the transaction flow data that needs to be reconciled into a pre-trained reconciliation model for matching analysis and processing, and obtain a reconciliation result;

[0189] The processing module is further configured to store the reconciliation result in a specific location in the blockchain when the reconciliation result indicates that the transaction flow data to be reconciled meets the consistency requirement;

[0190] The processing module is also used to generate a difference report and trigger an alarm mechanism when the reconciliation result shows that the transaction flow data that needs to be reconciled does not meet the consistency requirement.

[0191] It can be understood that the above-mentioned acquisition module and processing module can be used to execute the above-mentioned S11 to S14. The detailed implementation of the acquisition module and processing module can refer to the above-mentioned contents related to S11 to S14.

[0192] See also Figure 8 , is a block diagram of an electronic device provided in an embodiment of the present invention, and the electronic device may be a computer device, a server, etc., and may be any electronic device in a distributed architecture. The electronic device includes a memory, a processor, and a communication module. The memory, the processor, and the communication module are electrically connected to each other directly or indirectly to achieve data transmission or interaction. For example, these components may be electrically connected to each other via one or more communication buses or signal lines.

[0193] The memory is used to store programs or data. The memory may be, but is not limited to, random access memory (RAM), read only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable read-only memory (EEPROM), etc.

[0194] The processor is used to read / write data or programs stored in the memory and execute the automatic bank account reconciliation method provided by any embodiment of the present invention.

[0195] The communication module is used to establish a communication connection between the electronic device and other communication terminals through the network, and is used to send and receive data through the network.

[0196] It should be understood that Figure 8 The structure shown is only a schematic diagram of the structure of the electronic device. The electronic device may also include Figure 8 More or fewer components as shown, or with Figure 8 Different configurations are shown.

[0197] Furthermore, an embodiment of the present invention also provides a computer-readable storage medium, which stores machine-executable instructions. When the machine-executable instructions are executed, the automatic transaction reconciliation method provided in the above embodiment is implemented.

[0198] Specifically, the computer-readable storage medium can be a general storage medium, such as a mobile disk, a hard disk, etc. When the computer program on the computer-readable storage medium is executed, the above-mentioned automatic account reconciliation method can be executed. Regarding the process involved when the computer-readable storage medium and its executable instructions are executed, reference can be made to the relevant description in the above-mentioned method embodiment, which will not be described in detail here.

[0199] In the embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0200] In addition, the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0201] Furthermore, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0202] It should be noted that if the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0203] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0204] The above description is only an embodiment of the present invention and is not intended to limit the protection scope of the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An automatic account reconciliation method, characterized in that: The method comprises: Collecting transaction flow data, evaluating the transaction flow data to obtain an evaluation result, processing the transaction flow data based on the evaluation result, and submitting the transaction flow data to the blockchain for storage; In response to the reconciliation trigger condition, the transaction flow data that needs to be reconciled is extracted from the blockchain, and the transaction flow data that needs to be reconciled is input into the pre-trained reconciliation model for matching analysis and processing to obtain the reconciliation result; If the reconciliation result shows that the transaction flow data to be reconciled meets the consistency requirement, the reconciliation result is stored in a specific location in the blockchain; If the reconciliation result shows that the transaction flow data to be reconciled does not meet the consistency requirements, a difference report is generated and an alarm mechanism is triggered.

2. The automatic account reconciliation method according to claim 1, characterized in that: The step of evaluating the transaction flow data to obtain an evaluation result, and processing the transaction flow data based on the evaluation result includes: Evaluate the transaction flow data from multiple dimensions to obtain evaluation results of each dimension, wherein the multiple dimensions include completeness dimension, accuracy dimension, consistency dimension and timeliness dimension; Calculate the total evaluation value according to the evaluation results of each dimension, and filter out abnormal data in the transaction flow data according to the total evaluation value; The abnormal data is automatically corrected or marked.

3. The automatic account reconciliation method according to claim 1, characterized in that: The step of collecting transaction flow data includes: The collection tasks are allocated through the master control node, wherein the collection tasks are allocated according to any one of data source division, time range division and data volume division; The transaction flow data is collected by multiple working nodes according to the allocation of collection tasks, and the master control node records the execution status of each working node; The transaction flow data collected by each working node is stored through the data storage node.

4. The automatic account reconciliation method according to claim 1, characterized in that: The step of submitting the transaction flow data to the blockchain for storage includes: Grouping the transaction flow data in a set manner and generating a data fingerprint for each group of transaction flow data; Store each set of transaction flow data in the block body of each block of the blockchain in order, generate a hash value based on the block header and block body of the previous block, and store the hash value of the previous block in the block header of the next block; The full node in the blockchain is used to store the information of each block, and the light node in the blockchain is used to save the block header information of each block. The full node can be used to find the block where the required reconciliation transaction flow data is located.

5. The automatic account reconciliation method according to claim 4, characterized in that: Before the step of responding to the reconciliation trigger condition, the method further includes a step of verifying the transaction flow data in the blockchain, which step includes: Requesting the Merkle proof of the transaction flow data to be verified from the full node through the light node; Find the block where the transaction flow data is located through the full node, and construct a path from the transaction flow data to the Merkle root of the block, wherein the path from the transaction flow data to the Merkle root of the block is a Merkle proof; After receiving the Merkle proof, the light node uses the hash value in the Merkle proof to recalculate the Merkle root, and compares the calculated Merkle root with the Merkle root in the saved block header of the block. If the two are consistent, the verification is determined to be successful.

6. The automatic account reconciliation method according to claim 1, characterized in that: Before the step of inputting the transaction flow data to be reconciled into the pre-trained reconciliation model for matching analysis, the method further includes the step of pre-processing the transaction flow data to be reconciled, which step includes at least one of the following: Standardize the numerical features of transaction flow data that need to be reconciled; Performing feature coding processing on the transaction flow data; Converting the timestamp of the transaction flow data into a periodic time feature; An interaction feature is created based on the transaction flow data, wherein the interaction feature includes an amount difference and a time difference.

7. The automatic account reconciliation method according to claim 1, characterized in that: The reconciliation model includes a first sub-model and a second sub-model; The step of inputting the transaction flow data to be reconciled into the pre-trained reconciliation model for matching analysis and processing includes: Input the transaction flow data that needs to be reconciled into the first sub-model and the second sub-model that have been trained in advance respectively; Performing matching analysis and processing through the first sub-model to obtain a first reconciliation result, and performing matching analysis and processing through the second sub-model to obtain a second reconciliation result; The first reconciliation result and the second reconciliation result are weighted averaged to obtain the final reconciliation result.

8. The automatic account reconciliation method according to claim 1, characterized in that: After the step of collecting transaction flow data, the method further includes a step of identifying noise in the transaction flow data, which step includes at least one of the following: Using statistical methods to identify transaction flow data with abnormal values ​​in the transaction flow data; Transaction flow data that does not satisfy predefined business rules is identified in the transaction flow data.

9. The automatic account reconciliation method according to claim 1, characterized in that: After the step of acquiring transaction flow data, the method further includes a step of identifying abnormal values ​​in the transaction flow data, which step includes at least one of the following: Detecting outliers from the transaction flow data using statistical methods; The transaction flow data with abnormal fluctuations is detected by using a moving average detection method for the time series in the transaction flow data.

10. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images