A fund control method based on grassroots financial management
By obtaining and analyzing transaction flows in the flow documents, using semantic coding and abnormal detection technology, the problems of insufficient supervision and insensitive compliance detection in traditional fund control methods are solved, and more efficient fund monitoring and management are achieved.
Patent Information
- Application Number
- CN202411131865.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-19
AI Technical Summary
The traditional fund control methods in grassroots financial management lack comprehensive supervision and the compliance detection is not sensitive enough, resulting in unclear capital flows, regulatory loopholes and irregular fund management.
By obtaining flow files, extracting transaction flows, performing semantic coding and analysis, using important rule highlighters and classifiers based on class prospect attention mechanism for abnormal detection, generating abnormal flow alarms, and pushing alarms to discover abnormal transactions in a timely manner.
It improves the efficiency of monitoring capital flows, timely discovers abnormal transactions, optimizes the fund control process, and improves the accuracy and efficiency of fund control.
Smart Images

Figure CN118657618B_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the technical field of fund management and control, and in particular, to a fund management and control method based on grassroots financial management. Background Art
[0002] In grassroots financial management, financial departments often need to create, manage and allocate various budget items, and fund control is a crucial link. However, traditional fund control methods have some defects, resulting in insufficient supervision, insensitive compliance detection and other problems, which bring challenges to financial management. Specifically, traditional fund control solutions usually only check the information of the payee, lack of comprehensive supervision in the process of fund disbursement, and easily lead to problems such as unclear fund flow and regulatory loopholes. In addition, traditional methods lack an effective compliance detection mechanism in the fund disbursement link, or the compliance check is not sensitive enough, resulting in the inability to detect abnormal transactions or potential violations in time, which easily leads to irregular fund management. In addition, the current fund disbursement process is usually cumbersome and not very timely. The payment after approval cannot be completed in real time, and may need to be converted to offline payment, which increases the workload and risk.
[0003] Therefore, an optimized fund control plan based on grassroots financial management is desired. Summary of the invention
[0004] This summary is provided to introduce concepts in a brief form that will be described in detail in the detailed description below. This summary is not intended to identify key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.
[0005] In a first aspect, the present disclosure provides a method for fund control based on grassroots financial management, the method comprising:
[0006] Get the journal file;
[0007] Extracting transaction flow from the flow file;
[0008] Performing transaction flow analysis on the transaction flow to determine whether the transaction flow is abnormal;
[0009] In response to the transaction flow being abnormal, generating an abnormal flow alarm; and
[0010] Pushing the abnormal flow alarm;
[0011] Among them, the transaction flow is analyzed to determine whether the transaction flow is abnormal, including: extracting a set of reference flow data marked as normal flow from the background database; semantically encoding each reference flow data marked as normal flow in the set of reference flow data marked as normal flow to obtain a set of reference flow data semantic encoding feature vectors; arranging the set of reference flow data semantic encoding feature vectors into a full sample domain reference flow semantic synergy matrix and then using an important rule highlighter based on a foreground-like attention mechanism to obtain an enhanced full sample domain reference flow semantic synergy matrix; semantically encoding the transaction flow to obtain a detection transaction flow data semantic encoding feature vector; using the detection transaction flow data semantic encoding feature vector as a query feature vector to query a flow rule query feature from the enhanced full sample domain reference flow semantic synergy matrix; based on the flow rule query feature, determining whether the transaction flow is abnormal.
[0012] Optionally, the set of the reference pipeline data semantic encoding feature vectors is arranged into a reference pipeline semantic synergy matrix for the entire sample domain, and then an important rule highlighter based on a foreground-like attention mechanism is used to obtain an enhanced reference pipeline semantic synergy matrix for the entire sample domain, including: arranging the set of the reference pipeline data semantic encoding feature vectors in a matrix to obtain the reference pipeline semantic synergy matrix for the entire sample domain; using the negative of the eigenvalue of each position of the reference pipeline semantic synergy matrix for the entire sample domain as the exponent of a natural constant, calculating the exponential function value based on the position with the natural constant as the base, to obtain to the full sample domain reference pipeline semantics collaborative class support matrix; calculate the inverse of the sum of the eigenvalues of each position in the full sample domain reference pipeline semantics collaborative class support matrix and the constant one to obtain the full sample domain reference pipeline semantics collaborative representation matrix; mask the full sample domain reference pipeline semantics collaborative representation matrix to obtain the full sample domain reference pipeline semantics mask support weight matrix; perform positional point multiplication on the full sample domain reference pipeline semantics mask support weight matrix and the full sample domain reference pipeline semantics collaborative matrix to obtain the enhanced full sample domain reference pipeline semantics collaborative matrix.
[0013] Optionally, the full-sample-domain reference pipeline semantics co-representation matrix is masked to obtain a full-sample-domain reference pipeline semantics mask support weight matrix, including: setting the eigenvalues of each position in the full-sample-domain reference pipeline semantics co-representation matrix whose eigenvalues are less than a predetermined threshold to zero to obtain the full-sample-domain reference pipeline semantics mask support weight matrix.
[0014] Optionally, the semantically encoded feature vector of the detected transaction flow data is used as a query feature vector to query the flow rule query feature from the enhanced full sample domain reference flow semantic synergy matrix, including: using the semantically encoded feature vector of the detected transaction flow data as a query feature vector, calculating the matrix product between it and the enhanced full sample domain reference flow semantic synergy matrix to obtain the flow rule query feature vector as the flow rule query feature.
[0015] Optionally, determining whether the transaction flow is abnormal based on the flow rule query feature includes: passing the flow rule query feature vector through a classifier-based anomaly detector to obtain a detection result, wherein the detection result is used to indicate whether the transaction flow is abnormal.
[0016] Optionally, the method further includes a training step for training the important rule highlighter based on the class foreground attention mechanism and the classifier-based anomaly detector.
[0017] Optionally, the training step includes: obtaining a training flow file; extracting training transaction flow from the training flow file, as well as the true value of whether the transaction flow is abnormal; extracting a set of training reference flow data marked as normal flow from the background database; semantically encoding each training reference flow data marked as normal flow in the set of training reference flow data marked as normal flow to obtain a set of semantic encoding feature vectors of training reference flow data; arranging the set of semantic encoding feature vectors of training reference flow data into a training full sample domain reference flow semantic coordination matrix, and then using the important rule highlighter based on the class foreground attention mechanism to obtain a training enhanced full sample domain reference flow semantic coordination matrix. The same matrix; semantically encode the training transaction flow to obtain the semantically encoded feature vector of the training test transaction flow data; use the semantically encoded feature vector of the training test transaction flow data as the query feature vector, calculate the matrix product between it and the training enhanced full sample domain reference flow semantic synergy matrix to obtain the training flow rule query feature vector; pass the training flow rule query feature vector through the classifier-based anomaly detector to obtain the detection result; calculate the cross entropy loss function value between the detection result and the true result to obtain the classification loss function value; based on the classification loss function value, train the important rule highlighter based on the class foreground attention mechanism and the classifier-based anomaly detector.
[0018] The above technical solution is adopted to obtain transaction flow files; extract transaction flow from the transaction flow files; conduct transaction flow analysis to determine whether the transaction flow is abnormal; generate abnormal transaction flow alarm in response to the abnormal transaction flow; and push the transaction flow alarm. In this way, it can help the financial department to monitor the flow of funds more effectively and discover abnormal transactions in time, thereby helping to optimize the fund control process in grassroots financial management and improve the accuracy and efficiency of fund control.
[0019] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale. In the drawings:
[0021] Figure 1 It is a flow chart of a fund control method based on grassroots financial management according to an exemplary embodiment.
[0022] Figure 2 is based on Figure 1 The illustrated embodiment shows a flow chart of step S103 of a method for fund control based on grassroots financial management.
[0023] Figure 3 It is a block diagram of a fund control system based on grassroots financial management according to an exemplary embodiment.
[0024] Figure 4 The figure is a flowchart of setting a checkpoint for automatic inspection according to an exemplary embodiment.
[0025] Figure 5 The figure is a flow chart of pipeline data acquisition according to an exemplary embodiment.
[0026] Figure 6 The figure is a risk analysis flow chart according to an exemplary embodiment.
[0027] Figure 7 The following is a flowchart of executing project creation according to an exemplary embodiment.
[0028] Figure 8 The present invention is a flowchart of contract information registration and customer-merchant association according to an exemplary embodiment.
[0029] Fig. 9 The following is a flowchart of creating a payment application according to an exemplary embodiment.
[0030] Fig.10 The figure is a flowchart of approval flow approval according to an exemplary embodiment.
[0031] Fig.11 It is a block diagram of an electronic device according to an exemplary embodiment.
[0032] Fig.12 It is an application scenario diagram of a fund control method based on grassroots financial management according to an exemplary embodiment. DETAILED DESCRIPTION
[0033] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.
[0034] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.
[0035] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.
[0036] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.
[0037] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".
[0038] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.
[0039] Existing technologies and solutions have complicated processes and low timeliness in terms of fund disbursement. After the payment application is approved by the online system, it is no longer possible to complete the real-time payment online, and it may be transferred to offline, which increases the workload of the staff. In addition, various offline payment behaviors make it impossible to ensure the full supervision of the transaction flow of fund disbursement.
[0040] Similarly, due to various offline payment behaviors, existing technologies and solutions do not have corresponding compliance checks in the payment initiation process, or the compliance checks are not sensitive enough. It is difficult to track transaction flows in offline payments, and it is impossible to analyze transaction flow data through models in real time and finally issue alerts on transaction flows.
[0041] In order to solve the above problems, the present disclosure provides a fund control method based on grassroots financial management, which obtains a flow file; extracts transaction flow from the flow file; performs flow analysis on the transaction flow to determine whether the transaction flow is abnormal; generates an abnormal flow alarm in response to the abnormal transaction flow; and pushes the flow alarm. In this way, it can help the financial department to monitor the flow of funds more effectively and discover abnormal transactions in time, thereby helping to optimize the fund control process in grassroots financial management and improve the accuracy and efficiency of fund control.
[0042] The specific embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.
[0043] Figure 1 is a flow chart of a method for fund control based on grassroots financial management according to an exemplary embodiment. Figure 1 As shown, the method includes:
[0044] Step 101, obtaining a log file;
[0045] Step 102: extract transaction flow from the flow file;
[0046] Step 103: Analyze the transaction flow to determine whether the transaction flow is abnormal;
[0047] Step 104: In response to the transaction flow being abnormal, generating an abnormal flow alarm; and
[0048] Step 105: Push the flow alarm.
[0049] In response to the above technical problems, in the technical solution of the present application, a fund control method based on grassroots financial management is proposed, which includes: obtaining a transaction flow file; extracting transaction flow from the transaction flow file; performing flow analysis on the transaction flow to determine whether the transaction flow is abnormal; generating an abnormal flow alarm in response to the abnormal transaction flow; and pushing the flow alarm. The fund control method based on grassroots financial management can determine whether the transaction flow is abnormal through flow analysis, and generate an abnormal flow alarm when an abnormality or potential hidden danger is found. In this way, it can help the financial department to monitor the flow of funds more effectively and detect abnormal transactions in a timely manner, thereby helping to optimize the fund control process in grassroots financial management and improve the accuracy and efficiency of fund control.
[0050] Accordingly, in the above-mentioned fund control method based on grassroots financial management, the step of analyzing the transaction flow to determine whether the transaction flow is abnormal is particularly critical, which is an important step to ensure the normal transaction flow and optimize the accuracy and efficiency of fund control. Specifically, the technical concept of the present application is to extract a set of reference flow data marked as normal flow from the background database, and introduce semantic analysis and understanding algorithms in the back end to perform semantic analysis on these reference flow data, extract the full sample collaborative semantic representation information of the reference flow semantics, and query and compare it with the semantic coding representation information of the transaction flow, so as to determine whether the transaction flow is abnormal.
[0051] Figure 2 is based on Figure 1 The embodiment shown is a flow chart of step S103 of a method for fund control based on grassroots financial management. Figure 2 As shown, step 103, analyzing the transaction flow to determine whether the transaction flow is abnormal, includes:
[0052] Step 1031, extracting a set of reference flow data marked as normal flow from the background database;
[0053] Step 1032: semantically encode each reference pipeline data marked as normal pipeline in the set of reference pipeline data marked as normal pipeline to obtain a set of reference pipeline data semantic encoding feature vectors;
[0054] Step 1033, arranging the set of semantic encoding feature vectors of the reference pipeline data into a reference pipeline semantic synergy matrix for the entire sample domain, and then performing an important rule highlighter based on a foreground-like attention mechanism to obtain an enhanced reference pipeline semantic synergy matrix for the entire sample domain;
[0055] Step 1034: semantically encode the transaction flow to obtain a semantically encoded feature vector of the transaction flow data;
[0056] Step 1035: Use the semantic encoding feature vector of the detected transaction flow data as a query feature vector to query the flow rule query feature from the enhanced full sample domain reference flow semantic collaboration matrix;
[0057] Step 1036: Determine whether the transaction flow is abnormal based on the flow rule query characteristics.
[0058] Specifically, first, a set of reference flow data marked as normal flow is extracted from the background database. Next, considering that each reference flow data in the set of reference flow data marked as normal flow has flow patterns and features about normal transactions, in order to capture the normal flow semantics in these reference flow data, so as to more accurately identify abnormal transactions that do not conform to the normal pattern, in the technical solution of the present application, it is necessary to semantically encode each reference flow data marked as normal flow in the set of reference flow data marked as normal flow to obtain a set of semantically encoded feature vectors of reference flow data. By semantically encoding the normal flow data to obtain a set of feature vectors, the characteristics and patterns of normal transactions can be captured, and a reference standard can be established for subsequent comparative analysis of the detected transaction flow data. In this way, abnormal transactions can be better identified, because abnormal transactions are usually different from normal transactions, thereby improving the accuracy of abnormal flow detection.
[0059] Then, considering that when the transaction flow is actually detected for anomalies, there is a semantic association relationship of correlation between different normal transaction flow data, and at the same time, there may be some data that deviates from the normal transaction flow pattern. Therefore, in order to better capture the association and regularity between the semantics of different normal transaction flow data, the semantic feature information of all reference flow data is comprehensively considered, and the important rules and semantics therein are highlighted, so as to facilitate the subsequent identification of transaction flow anomalies and improve the sensitivity and accuracy of anomaly detection. In the technical solution of the present application, the set of the semantic encoding feature vectors of the reference flow data is further arranged into a reference flow semantic synergy matrix of the whole sample domain, and then an important rule highlighter based on the class foreground attention mechanism is used to obtain an enhanced reference flow semantic synergy matrix of the whole sample domain. Through the processing of the important rule highlighter based on the class foreground attention mechanism, the reference flow semantic synergy association representation features based on the whole sample domain between the semantics of each normal transaction flow data can be learned and captured, and the important rules and features are highlighted, so that the system can pay more attention to those parts that may contain abnormal transaction information when performing abnormal flow detection in the future, thereby improving the efficiency and accuracy of anomaly detection.
[0060] In one embodiment of the present disclosure, the set of semantically encoded feature vectors of the reference pipeline data is arranged into a reference pipeline semantic synergy matrix for the entire sample domain, and then an important rule highlighter based on a foreground-like attention mechanism is used to obtain an enhanced reference pipeline semantic synergy matrix for the entire sample domain, including: arranging the set of semantically encoded feature vectors of the reference pipeline data in a matrix to obtain the reference pipeline semantic synergy matrix for the entire sample domain; using the negative of the feature value of each position of the reference pipeline semantic synergy matrix for the entire sample domain as the exponent of the natural constant, and calculating the exponential function based on the position with the natural constant as the base value to obtain the reference pipeline semantics collaborative class support matrix for the entire sample domain; calculate the inverse of the sum of the eigenvalues of each position in the reference pipeline semantics collaborative class support matrix for the entire sample domain and the constant one to obtain the reference pipeline semantics collaborative representation matrix for the entire sample domain; mask the reference pipeline semantics collaborative representation matrix for the entire sample domain to obtain the reference pipeline semantics mask support weight matrix for the entire sample domain; perform positional point multiplication on the reference pipeline semantics mask support weight matrix for the entire sample domain and the reference pipeline semantics collaborative matrix for the entire sample domain to obtain the enhanced reference pipeline semantics collaborative matrix for the entire sample domain.
[0061] Furthermore, the full-sample-domain reference pipeline semantics collaborative representation matrix is masked to obtain a full-sample-domain reference pipeline semantics mask support weight matrix, including: setting the eigenvalues of each position in the full-sample-domain reference pipeline semantics collaborative representation matrix whose eigenvalues are less than the predetermined threshold to zero to obtain the full-sample-domain reference pipeline semantics mask support weight matrix.
[0062] Furthermore, the transaction flow is further semantically encoded to obtain a semantically encoded feature vector for detecting the transaction flow data. Semantic features in the transaction flow data can be extracted through semantic encoding and converted into semantically encoded feature vectors. These feature vectors can better describe the meaning and characteristics of the transaction flow data. In this way, the system can use these feature vectors to perform anomaly detection and identify transactions that may have abnormalities, such as abnormal amounts, abnormal times, etc., so as to promptly discover potential problems, improve the accuracy and efficiency of anomaly detection, and reduce false positives and false negatives.
[0063] Further, considering that the enhanced full-sample domain reference flow semantic synergy matrix contains the semantic features and patterns of the normal transaction flow set after the foreground content is highlighted, and the semantic coding feature vector of the detected transaction flow data represents the semantic feature information of the detected transaction flow data, therefore, in order to be able to judge whether the semantics of the detected transaction flow data conform to the pattern of normal transaction flow, thereby improving the efficiency and accuracy of abnormal detection, in the technical solution of the present application, the semantic coding feature vector of the detected transaction flow data is further used as a query feature vector, and the matrix product between it and the enhanced full-sample domain reference flow semantic synergy matrix is calculated to obtain the flow rule query feature vector. By using the semantic coding feature vector of the detected transaction flow data as a query feature vector, it can be used as a query factor to query the relevant pattern features from the enhanced full-sample domain reference flow semantic synergy matrix, so as to judge whether the detected transaction flow data conforms to the normal transaction flow pattern. In this way, it is easier to discover the potential laws between the semantics of normal transaction flow data, so as to more effectively discover the possible abnormal transaction situation, and help the financial management department to take timely measures to deal with potential risks.
[0064] In one embodiment of the present disclosure, the semantically encoded feature vector of the detected transaction flow data is used as a query feature vector to query the flow rule query feature from the enhanced full sample domain reference flow semantic synergy matrix, including: using the semantically encoded feature vector of the detected transaction flow data as a query feature vector, calculating the matrix product between it and the enhanced full sample domain reference flow semantic synergy matrix to obtain the flow rule query feature vector as the flow rule query feature.
[0065] After arranging the set of the semantic encoding feature vectors of the reference pipeline data into a reference pipeline semantic synergy matrix for the entire sample domain, the matrix is processed by the important rule highlighter based on the foreground attention mechanism using the following foreground explicitation formula to obtain the enhanced reference pipeline semantic synergy matrix for the entire sample domain;
[0066] The prospect manifestation formula is:
[0067]
[0068] in, is the first part of the reference pipeline semantic synergy matrix of the whole sample domain The eigenvalues of the positions, is the exponential operation, for function, The first part of the weight matrix of the reference pipeline semantic collaborative mask support for the full sample domain The eigenvalues of the positions, is a hyperparameter, is the reference pipeline semantic synergy matrix of the whole sample domain, is the reference pipeline semantic collaborative mask support weight matrix for the full sample domain, For point-by-position multiplication, The enhanced full-sample domain reference flow semantic coordination matrix. Then, the flow rule query feature vector is passed through a classifier-based anomaly detector to obtain a detection result, and the detection result is used to indicate whether the transaction flow is abnormal. In other words, the query feature vector is mapped to the flow rule query feature in the enhanced full-sample domain reference flow semantic coordination matrix for classification processing, so as to identify and judge whether the transaction flow is abnormal, and generate an abnormal flow alarm when an anomaly or potential hidden danger is found. In this way, it can help the financial department to monitor the flow of funds more effectively and detect abnormal transactions in a timely manner, thereby helping to optimize the fund control process in grassroots financial management and improve the accuracy and efficiency of fund control.
[0069] In one embodiment of the present disclosure, determining whether a transaction flow is abnormal based on the flow rule query feature includes: passing the flow rule query feature vector through a classifier-based anomaly detector to obtain a detection result, wherein the detection result is used to indicate whether the transaction flow is abnormal.
[0070] In particular, in the technical solution of the present application, the semantic encoding feature vector of the detected transaction flow data represents the semantic feature representation of the transaction flow data based on the semantic encoder, and the enhanced full-sample domain reference flow semantic synergy matrix represents the important feature saliency representation based on the class foreground attention mechanism of the full-sample domain semantic association of the set of reference flow data marked as normal flow. Considering that the semantic encoding feature vector of the detected transaction flow data is obtained by single-sample semantic encoding, and the enhanced full-sample domain reference flow semantic synergy matrix is obtained by multi-sample domain and attention saliency, therefore, when the semantic encoding feature vector of the detected transaction flow data is used as the query feature vector and the matrix product between it and the enhanced full-sample domain reference flow semantic synergy matrix is calculated, the obtained flow rule query feature vector has insufficient interactive coverage caused by the different feature distribution and feature distribution order difference between the single-sample domain and the multi-sample domain, thereby causing the outlier class regression inference mapping deviation, affecting the accuracy of the detection result obtained by the flow rule query feature vector input to the classifier-based anomaly detector.
[0071] Based on this, when the pipeline rule query feature vector is input into the classifier-based anomaly detector to obtain the detection result, the pipeline rule query feature vector is optimized, and the optimization process specifically includes:
[0072] Calculating a feature mean of the pipeline rule query feature vector, and dividing the feature mean by a difference between a maximum feature value and a minimum feature value of the pipeline rule query feature vector to obtain a pipeline rule query feature distribution representation value;
[0073] Subtracting the pipeline rule query characteristic distribution representation value from one and dividing the result by the pipeline rule query characteristic distribution representation value to obtain a pipeline rule query characteristic distribution modulation value;
[0074] Activating the pipeline rule query feature vector through a probabilistic function to obtain a probabilistic pipeline rule query feature vector;
[0075] After performing point subtraction between the probabilistic pipeline rule query feature vector and the pipeline rule query feature distribution modulation value, taking the absolute value and calculating the negative of the logarithmic value with base 2 to obtain the probabilistic pipeline rule query feature distribution modulation information vector;
[0076] After dividing the pipeline rule query feature distribution characterization value by one minus the difference of each feature value of the probabilistic pipeline rule query feature vector, summing all feature values of the probabilistic pipeline rule query feature vector and dividing by the length of the pipeline rule query feature vector to obtain a probabilistic pipeline rule query feature distribution modulation bias value;
[0077] The product of the probabilistic pipeline rule query feature distribution modulation information vector and the probabilistic pipeline rule query feature distribution modulation bias value and the weight as a hyperparameter is multiplied to obtain an optimized pipeline rule query feature vector.
[0078] The optimized expression of the pipeline rule query feature vector is as follows:
[0079]
[0080] in, is the probabilistic pipeline rule query feature distribution representation value, The middle value representing the probabilistic pipeline rule query feature distribution representation value eigenvalues, is the pipeline rule query characteristic distribution modulation value, is the length of the distribution representation value of the pipeline rule query feature, is the weight hyperparameter, It means to make a difference by position. It means adding by position. is the optimized pipeline rule query feature distribution representation value; and wherein:
[0081]
[0082] in, represents the feature mean of the distribution representation value of the pipeline rule query feature, and respectively represent the maximum eigenvalue and the minimum eigenvalue of the characteristic distribution representation value of the pipeline rule query, is the time series aggregate distribution representation value of the lip movement feature. That is, in the above preferred example, the probability information distribution planning based on the feature value of the pipeline rule query feature vector is carried out through the Bernoulli probability modulation distribution of the pipeline rule query feature vector relative to the feature value distribution, and the probability reverse mapping of the probability feature of the pipeline rule query feature vector as a whole is used as an extended coverage of the set mapping space of the pipeline rule query feature vector to independently understand the interaction path between the intuitive probability information distribution and the abstract probability space mapping of the pipeline rule query feature vector, so as to improve the accuracy of the detection result obtained by the optimized pipeline rule query feature vector input based on the classifier anomaly detector by avoiding the counterfactual reasoning mapping of the outlier feature distribution of the pipeline rule query feature vector to the class regression probability.
[0083] Furthermore, in one embodiment of the present disclosure, the fund control method based on grassroots financial management also includes a training step: for training the important rule highlighter based on the class foreground attention mechanism and the classifier-based anomaly detector. The training step includes: obtaining a training flow file; extracting training transaction flow from the training flow file, as well as the true value of whether the transaction flow is abnormal; extracting a set of training reference flow data marked as normal flow from the background database; semantically encoding each training reference flow data marked as normal flow in the set of training reference flow data marked as normal flow to obtain a set of training reference flow data semantic encoding feature vectors; arranging the set of training reference flow data semantic encoding feature vectors into a training full sample domain reference flow semantic synergy matrix and then passing it through the important rule highlighter based on the class foreground attention mechanism to obtain a training enhanced full sample domain reference flow semantic synergy matrix. ; Perform semantic encoding on the training transaction flow to obtain a semantic encoding feature vector of the training test transaction flow data; use the semantic encoding feature vector of the training test transaction flow data as a query feature vector, calculate the matrix product between it and the training enhanced full sample domain reference flow semantic synergy matrix to obtain a training flow rule query feature vector; pass the training flow rule query feature vector through the classifier-based anomaly detector to obtain a detection result; calculate the cross entropy loss function value between the detection result and the true result to obtain a classification loss function value; train the important rule highlighter based on the class foreground attention mechanism and the classifier-based anomaly detector based on the classification loss function value.
[0084] In a preferred embodiment of the above technical solution, the training flow rule query feature vector is passed through a classifier-based anomaly detector to obtain a detection result, and the training detection result is used to indicate whether the transaction flow is abnormal, specifically including:
[0085] Dividing each eigenvalue of the training pipeline rule query feature vector by the maximum eigenvalue of the training pipeline rule query feature vector to obtain a training pipeline rule query semantic interaction representation vector;
[0086] Dividing the eigenvalue mean of the training pipeline rule query feature vector by the eigenvalue standard deviation of the training pipeline rule query feature vector to obtain a statistical dimension interaction value corresponding to the training pipeline rule query feature vector;
[0087] After subtracting the statistical dimension interaction value from each eigenvalue of the training pipeline rule query semantic interaction representation vector, calculate the logarithm of each position thereof to obtain the training pipeline rule query semantic interaction information representation vector;
[0088] Adding each eigenvalue of the training pipeline rule query semantic interaction representation vector to the statistical dimension interaction value and then multiplying by a predetermined weight hyperparameter to obtain a training pipeline rule query semantic interaction pattern representation vector;
[0089] The training pipeline rule query semantic interaction information representation vector is added to the training pipeline rule query semantic interaction pattern representation vector to obtain an optimized pipeline rule query feature vector; and
[0090] The optimized pipeline rule query feature vector is passed through the classifier-based anomaly detector to obtain the detection result.
[0091] That is, in order to improve the feature expression effect of the training pipeline rule query feature vector in the dimension of complex feature association representation of heterogeneous feature distribution and feature distribution order interaction in single sample domain-multiple sample domain, in the above steps, a short sequence of dimensional interaction representation of statistical features represented by mean and standard value and distribution interaction representation of eigenvalue and maximum eigenvalue as latent variable features is used as a sub-manifold potential motif under the complex manifold network of the training pipeline rule query feature vector, so as to obtain the training pipeline rule query feature vector based on the training pipeline rule query semantic interaction information representation vector and the training pipeline rule query semantic The latent motif feature information pattern and feature distribution pattern of the interactive pattern representation vector are used as its global structure inference unit, so as to reconstruct the complex manifold structure of the training flow rule query feature vector in the form of a global structure latent motif dictionary based on connection, so as to improve the model's ability to generate and evolve the manifold structure corresponding to the features under the complex feature representation dimension during the iteration process, and improve the feature expression effect of the complex feature representation dimension of the training flow rule query feature vector during the model iteration process, so as to improve the accuracy of the detection result obtained by the training flow rule query feature vector through the classifier-based anomaly detector. In this way, it is possible to more accurately detect whether the transaction flow is abnormal, and generate an abnormal flow alarm when anomalies or potential hidden dangers are found, thereby helping the financial department to monitor the flow of funds more effectively, helping to optimize the fund control process in grassroots financial management, and improving the accuracy and efficiency of fund control.
[0092] To summarize, the above scheme is adopted to extract a set of reference transaction data marked as normal transactions from the background database, and introduce semantic analysis and understanding algorithms in the back end to perform semantic analysis on these reference transaction data, from which the full-sample collaborative semantic representation information of the reference transaction semantics is extracted, and queried and compared with the semantic coding representation information of the transaction transaction, so as to determine whether the transaction transaction is abnormal.
[0093] Figure 3 FIG. 1 is a block diagram of a fund control system based on grassroots financial management according to an exemplary embodiment. Figure 3 As shown, the fund control system 200 includes:
[0094] A pipeline file acquisition module 201 is used to acquire a pipeline file;
[0095] A transaction flow extraction module 202, used to extract transaction flow from the flow file;
[0096] A transaction flow analysis module 203, configured to perform transaction flow analysis on the transaction flow to determine whether the transaction flow is abnormal;
[0097] The abnormal transaction flow alarm generating module 204 is used to generate an abnormal transaction flow alarm in response to the abnormal transaction flow; and
[0098] The flow alarm push module 205 is used to push the flow alarm.
[0099] In one embodiment of the present disclosure, Figure 4 As shown in the figure, during the approval and execution of business documents (payment applications), the fund control system will set checkpoints to automatically check the attachments, invoices and other information submitted by the business documents. After the payment is completed, sensitive transactions will be monitored based on the transaction flow.
[0100] Figure 5 It is the process of acquiring transaction data. The fund control system asynchronously acquires large quantities of transaction data and parses and stores them in the database. The transaction data will be integrated according to the information of the payer, payee, etc., and the system will perform risk analysis.
[0101] like Figure 6 As shown in the figure, the transaction information is used as a multidimensional data set, and the outliers in the data set are identified through the LOF outlier detection algorithm. The outliers are determined by measuring the relative degree of abnormality of each data point and its surrounding neighboring points. Alarms are pushed for the transaction flow of the outliers.
[0102] The disclosure realizes a fully digital process, and the payment of various pending budgets can be carried out online in real time, making the payment behavior traceable and trackable, avoiding the inconvenience and risks brought by offline operations. This technology reduces the violations in business operations by adding compliance checks in the process of project creation, contract registration, invoice upload, document information completion, and fund payment. This technology uses digital models to conduct multi-dimensional analysis of subsequent transaction flows, warns of risky transaction behaviors that have already occurred, and helps users identify potential problems.
[0103] The business process involves business data creation, process approval, payment, etc. Business personnel first need to create an execution project and fill in detailed information such as the contract and payee. After allocating the corresponding funding indicators through the execution project, they can initiate a payment application.
[0104] Figure 7 To create an execution project, the specific execution project includes multiple types and creation methods.
[0105] Figure 8 To register contract information and associate with merchants, users need to log in to the merchant information, associate it with the specific contract, and complete the relevant elements of the contract.
[0106] Fig. 9It is used to create a payment application. After the payment application is created, a payment order will be generated, and the payment order can be paid. When applying for payment, you can associate a contract and select the corresponding indicators. The necessary supplementary documents and attachments for the payment application can be submitted for review by auditors and system automated compliance checks. The payment order supports multiple payment methods, communicates with the bank's online banking interface in the form of encrypted messages, and performs idempotent processing.
[0107] Fig.10 In the process of creating business objects such as execution projects, funding indicators, and payment applications, they all need to go through the approval process and be confirmed by the reviewers of the relevant nodes (electronic signatures can be recorded).
[0108] Reference below Fig.11 , which shows a schematic diagram of the structure of an electronic device 600 suitable for implementing the embodiment of the present disclosure. The terminal device in the embodiment of the present disclosure may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle terminals (such as vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Fig.11 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.
[0109] like Fig.11 As shown, the electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.
[0110] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Fig.11 The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.
[0111] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.
[0112] It should be noted that the computer-readable medium mentioned above in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.
[0113] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.
[0114] The computer-readable medium may be included in the electronic device, or may exist independently without being installed in the electronic device.
[0115] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages or a combination thereof, including, but not limited to, object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0116] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
[0117] The modules involved in the embodiments described in the present disclosure may be implemented by software or hardware. The name of a module does not limit the module itself in some cases. For example, a test parameter acquisition module may also be described as a "module for acquiring device test parameters corresponding to a target device".
[0118] Fig.12 FIG. 1 is an application scenario diagram of a method for controlling funds based on grassroots financial management according to an exemplary embodiment. Fig.12 As shown, in this application scenario, first, a set of reference flow data marked as normal flow is extracted from the background database (for example, Fig.12 C1 shown in the figure); extracting transaction flow from the flow file (for example, Fig.12 Then, the obtained reference flow data set and transaction flow are input into a server that is deployed with a fund control algorithm based on grassroots financial management (for example, Fig.12 In S) shown in , the server is capable of processing the set of reference flow data and the transaction flow based on the fund control algorithm of grassroots financial management to determine whether the transaction flow is abnormal.
[0119] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.
[0120] In addition, although each operation is described in a specific order, this should not be understood as requiring these operations to be performed in the specific order shown or in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.
[0121] Although the subject matter has been described in language specific to structural features and / or method logic actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims. Regarding the device in the above embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.
Claims
1. A fund control method based on grassroots financial management, characterized in that: include: Get the journal file; Extracting transaction flow from the flow file; Performing transaction flow analysis on the transaction flow to determine whether the transaction flow is abnormal; In response to the transaction flow being abnormal, generating an abnormal flow alarm; as well as Pushing the abnormal flow alarm; The step of analyzing the transaction flow to determine whether the transaction flow is abnormal includes: Extracting a set of reference flow data marked as normal flow from a backend database; Performing semantic coding on each reference pipeline data marked as normal pipeline in the set of reference pipeline data marked as normal pipeline to obtain a set of reference pipeline data semantic coding feature vectors; Arranging the set of semantic encoding feature vectors of the reference pipeline data into a reference pipeline semantic synergy matrix for the entire sample domain, and then using an important rule highlighter based on a foreground-like attention mechanism to obtain an enhanced reference pipeline semantic synergy matrix for the entire sample domain; Semantically encoding the transaction flow to obtain a semantically encoded feature vector of the transaction flow data; Using the semantic encoding feature vector of the detected transaction flow data as a query feature vector to query the flow rule query feature from the enhanced full sample domain reference flow semantic collaboration matrix; Determine whether the transaction flow is abnormal based on the transaction rule query characteristics; After arranging the set of semantic encoding feature vectors of the reference pipeline data into a reference pipeline semantic synergy matrix for the entire sample domain, an important rule highlighter based on a foreground-like attention mechanism is used to obtain an enhanced reference pipeline semantic synergy matrix for the entire sample domain, including: Arranging the set of the reference pipeline data semantic encoding feature vectors in a matrix to obtain the full sample domain reference pipeline semantic coordination matrix; Taking the negative number of the eigenvalue of each position of the reference pipeline semantic synergy matrix of the whole sample domain as the exponent of the natural constant, calculating the exponential function value with the natural constant as the base according to the position, so as to obtain the reference pipeline semantic synergy class support matrix of the whole sample domain; Calculating the inverse of the sum of the eigenvalues of each position in the full sample domain reference pipeline semantics collaborative class support matrix and a constant one to obtain a full sample domain reference pipeline semantics collaborative representation matrix; Masking the full-sample domain reference pipeline semantics collaborative representation matrix to obtain a full-sample domain reference pipeline semantics mask support weight matrix; The full-sample-domain reference pipeline semantic mask support weight matrix and the full-sample-domain reference pipeline semantic synergy matrix are point-wise multiplied to obtain the enhanced full-sample-domain reference pipeline semantic synergy matrix.
2. The method for fund control based on grassroots financial management according to claim 1, characterized in that: The full-sample-domain reference pipeline semantics collaborative representation matrix is masked to obtain the full-sample-domain reference pipeline semantics mask support weight matrix, including: setting the eigenvalues of each position in the full-sample-domain reference pipeline semantics collaborative representation matrix whose eigenvalues are less than a predetermined threshold to zero to obtain the full-sample-domain reference pipeline semantics mask support weight matrix.
3. The method for fund control based on grassroots financial management according to claim 2, characterized in that: The semantic encoding feature vector of the detected transaction flow data is used as a query feature vector to query the flow rule query feature from the enhanced full sample domain reference flow semantic synergy matrix, including: using the semantic encoding feature vector of the detected transaction flow data as a query feature vector, calculating the matrix product between it and the enhanced full sample domain reference flow semantic synergy matrix to obtain the flow rule query feature vector as the flow rule query feature.
4. The method for fund control based on grassroots financial management according to claim 3 is characterized in that: Based on the transaction rule query feature, determining whether the transaction flow is abnormal includes: passing the transaction rule query feature vector through a classifier-based anomaly detector to obtain a detection result, wherein the detection result is used to indicate whether the transaction flow is abnormal.
5. The method for fund control based on grassroots financial management according to claim 4 is characterized in that: It also includes a training step: used to train the important rule highlighter based on the class foreground attention mechanism and the classifier-based anomaly detector.
6. The method for fund control based on grassroots financial management according to claim 5, characterized in that: The training step comprises: Get the training pipeline file; Extracting training transaction flow from the training flow file, and a true value of whether the transaction flow is abnormal; Extracting a set of training reference pipeline data marked as normal pipelines from a backend database; Performing semantic coding on each of the training reference pipeline data marked as normal pipeline in the set of training reference pipeline data marked as normal pipeline to obtain a set of semantic coding feature vectors of the training reference pipeline data; Arranging the set of semantic encoding feature vectors of the training reference pipeline data into a training full sample domain reference pipeline semantic synergy matrix and then passing it through the important rule highlighter based on the foreground-like attention mechanism to obtain a training enhanced full sample domain reference pipeline semantic synergy matrix; Semantically encoding the training transaction flow to obtain a semantically encoded feature vector of the training test transaction flow data; Taking the semantic encoding feature vector of the training detection transaction flow data as the query feature vector, calculating the matrix product between the feature vector and the training enhanced full sample domain reference flow semantic synergy matrix to obtain the training flow rule query feature vector; Passing the training pipeline rule query feature vector through the classifier-based anomaly detector to obtain a detection result; Calculating a cross entropy loss function value between the detection result and the true result to obtain a classification loss function value; The important rule highlighter based on the class foreground attention mechanism and the classifier-based anomaly detector are trained based on the classification loss function value.
