Key account analysis method and system based on fund transaction

Abstract

The invention provides a key account analysis method and system based on fund transaction. A data acquisition and processing module is used for access and preprocessing of fund transaction data; the key account analysis module is used for performing key account analysis processing on the preprocessed fund transaction data by using a k-truss algorithm, and automatically marking a key account to obtain a key account analysis result; and the visual display module is used for displaying the key account analysis result. According to the method, large-scale fund transaction data is processed, the transaction relation graph is constructed, and the key accounts in the relation graph are identified based on the k-trus algorithm, so that the identification efficiency is better and more accurate, different analysis requirements are met, and the capability of financial institutions in fund flow monitoring and risk management aspects is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of financial data processing, and in particular to a key account analysis method and system based on capital transactions. Background Art

[0002] With the rapid development of the financial market, capital flows are becoming increasingly frequent and complex. Effectively monitoring capital flows and identifying key accounts are crucial for anti-money laundering, anti-fraud and compliance management. The existing processing method generally involves first cleaning the capital transaction data with data cleaning tools, and then using manual analysis methods or algorithms based on knowledge graphs to discover key accounts in capital transactions. This has problems such as low efficiency in large-scale data processing and low accuracy in identifying key accounts.

[0003] As the amount of fund transaction data increases, the existing distance discovery or depth-breadth search algorithms based on knowledge graphs and relationship graphs are used to identify key accounts in the transaction network, which has the problem of significantly increased computational complexity. In addition, these methods are sensitive to the selection of initial parameters. When the transaction network graph data changes dynamically (such as the addition of new nodes or edges), recalculation is required, resulting in low computational efficiency. In addition, these methods for calculating key accounts are generally based on known single or multiple account information. Distance discovery or depth-breadth search algorithms are used to obtain relevant key account information in the network through these known node information. It is impossible to perform an overall analysis of the nodes in the transaction network, and it is impossible to obtain key accounts in the entire transaction network by adjusting the community's closeness parameters to adapt to different analysis needs.

[0004] Therefore, it is urgent to develop an efficient and accurate key account analysis method and system. Summary of the invention

[0005] Based on the above content, the present invention provides a key account analysis method and system based on financial transactions, aiming to solve the technical problems of low efficiency, low accuracy and limited application scope of the key account identification method in the prior art.

[0006] A key account analysis system based on fund transactions, comprising:

[0007] Data collection and processing module, used for access and preprocessing of fund transaction data;

[0008] The key account analysis module is connected to the data acquisition and processing module and is used to use the k-truss algorithm to perform key account analysis on the pre-processed fund transaction data, automatically mark key accounts, and obtain key account analysis results;

[0009] The visualization display module is connected to the key account analysis module to display the key account analysis results.

[0010] Furthermore, the data acquisition and processing module includes:

[0011] A data access unit, used to access fund transaction data;

[0012] A data cleaning unit, connected to the data access unit, is used to clean the fund transaction data according to preset cleaning rules;

[0013] The data conversion unit is connected to the data cleaning unit and is used to convert the cleaned fund transaction data according to preset data standardization rules to obtain standardized fund transaction data.

[0014] Furthermore, the key account analysis module includes:

[0015] A grouping statistics unit is used to group and count the fund transaction data based on the local account and the peer account to form aggregated transaction data;

[0016] A relationship graph construction unit is connected to the grouping statistics unit and is used to use the local account and the peer account in the aggregated transaction data as nodes and the transaction relationship between the local account and the peer account as edges to construct and save the transaction relationship graph;

[0017] The key account identification unit and the connection relationship graph construction unit are used to calculate the support of the edges in the transaction relationship graph using the k-truss algorithm, screen the nodes and edges according to the support of the edges to construct a k-truss subgraph, calculate the key nodes from the k-truss subgraph as the key accounts, and form the key account analysis results.

[0018] Furthermore, the visualization module includes:

[0019] A relationship graph query unit, used for a user to input query conditions, the query conditions including a support threshold;

[0020] The key account identification unit is used to:

[0021] The transaction relationship subgraph is initially filtered according to the query conditions input by the user to form a k-truss subgraph. The initial filtering includes filtering out the edges and nodes whose edge support is greater than the support threshold to form a k-truss subgraph;

[0022] Calculate the key nodes from the k-truss subgraph as key accounts to form key account analysis results;

[0023] Return the key account analysis results shown to the visualization display module;

[0024] The visualization module also includes:

[0025] Visual interaction unit, connection relationship graph query unit, used to display the searched key account analysis results.

[0026] Furthermore, the attributes of the nodes in the transaction relationship graph include account number and account holder name;

[0027] The attributes of the edges in the transaction relationship graph include transaction frequency and transaction amount;

[0028] The query condition input by the user also includes at least one of transaction amount, transaction frequency, account number, and account holder name.

[0029] A key account analysis method based on capital transactions, using the aforementioned key account analysis system based on capital transactions, comprises:

[0030] Step A1, accessing fund transaction data and preprocessing;

[0031] Step A2, using the k-truss algorithm to perform key account analysis on the pre-processed fund transaction data, automatically marking key accounts, and obtaining key account analysis results;

[0032] Step A3: Display the key account analysis results.

[0033] Further, step A1 comprises:

[0034] Step A11, accessing fund transaction data;

[0035] Step A12, cleaning the fund transaction data according to preset cleaning rules;

[0036] Step A13, converting the cleaned fund transaction data according to preset data standardization rules to obtain standardized fund transaction data.

[0037] Further, step A2 comprises:

[0038] Step A21, grouping and counting the fund transaction data based on the local account and the peer account to form aggregated transaction data;

[0039] Step A22, using the local account and the peer account in the aggregated transaction data as nodes, and the transaction relationship between the local account and the peer account as an edge, to construct and save a transaction relationship graph;

[0040] Step A23, use the k-truss algorithm to calculate the support of the edges in the transaction relationship graph, screen the nodes and edges according to the support of the edges to construct a k-truss subgraph, calculate the key nodes from the k-truss subgraph as key accounts, and form the key account analysis results.

[0041] Furthermore, before step A23, step B1 is also included: the user inputs a query condition, and the query condition includes a support threshold;

[0042] In step A23, the transaction relationship subgraph is initially filtered according to the query conditions input by the user to form a k-truss subgraph, and the key nodes are calculated from the k-truss subgraph as key accounts to form and return the key account analysis results. The initial filtering includes screening out the edges and nodes whose edge support is greater than the support threshold.

[0043] Further, in step A22, the attributes of the nodes in the transaction relationship graph include the account number and the name of the account holder, and the attributes of the edges in the transaction relationship graph include the transaction frequency and the transaction amount;

[0044] In step B1, the query condition input by the user also includes at least one of transaction amount, transaction frequency, account number, and account holder name.

[0045] The beneficial technical effect of the present invention is that the present invention processes large-scale fund transaction data, constructs a transaction relationship graph, and identifies key accounts in the relationship graph based on the k-truss algorithm, with better identification efficiency, accuracy, and real-time performance. Moreover, the overall key account can be identified, rather than just limited to a single account or multiple accounts, so as to adapt to different analysis needs and help improve the ability of financial institutions in fund flow monitoring and risk management. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a module schematic diagram of a key account analysis system based on fund transactions of the present invention;

[0047] Figure 2-4 The present invention is a flowchart of the steps of a key account analysis method based on fund transactions. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0050] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0051] See also Figure 1 The present invention provides a key account analysis system based on fund transactions, comprising:

[0052] Data collection and processing module (1), used for accessing and preprocessing fund transaction data;

[0053] The key account analysis module (2) is connected to the data collection and processing module (1) and is used to perform key account analysis on the pre-processed fund transaction data using a k-truss algorithm, automatically mark key accounts, and obtain key account analysis results;

[0054] The visualization display module (3) is connected to the key account analysis module (2) and is used to display the key account analysis results.

[0055] The present invention processes large-scale fund transaction data, constructs a transaction relationship graph, and identifies key accounts in the relationship graph based on the k-truss algorithm, with better recognition efficiency, accuracy, and real-time performance. Moreover, the overall key account can be identified, rather than just limited to a single account or multiple accounts, so as to adapt to different analysis needs and help improve the ability of financial institutions in fund flow monitoring and risk management.

[0056] Furthermore, the data acquisition and processing module (1) comprises:

[0057] A data access unit (11), used for accessing fund transaction data;

[0058] A data cleaning unit (12), connected to the data access unit (11), is used to clean the fund transaction data according to a preset cleaning rule;

[0059] The data conversion unit (13) is connected to the data cleaning unit (12) and is used to convert the cleaned fund transaction data according to preset data standardization rules to obtain standardized fund transaction data.

[0060] The data access unit realizes the access of fund transaction data, supports real-time or batch data docking with other information systems through API interface or database connection, and supports the import of transaction data files in various formats (such as CSV, Excel, JSON, etc.) to process historical transaction records.

[0061] The preset cleaning rules include anomaly judgment rules and data deduplication rules. The data cleaning unit identifies and deletes duplicate records in the received fund transaction data according to the configured preset cleaning rules to ensure the uniqueness and accuracy of the data set. Detect missing values ​​in key fields in the data and estimate missing values ​​through account information or interpolation methods. Identify outliers in fund transaction data and delete them or modify or replace them according to the context.

[0062] Specifically, the deduplication fields of the data deduplication rules in data cleaning include the local account, the peer account, the loan flag, the transaction amount, and the transaction time. For the processing of missing values, you can use a fixed value, mean, median, mode, or estimate through a prediction model, or process the data in time series to estimate the missing values ​​based on the values ​​of adjacent data points. According to the anomaly judgment rules, determine whether the data record is a true outlier. The outlier processing strategy includes deleting, correcting, or replacing the outlier.

[0063] The data conversion unit is mainly responsible for normalizing the access data content according to preset data standardization rules to facilitate subsequent key account analysis. For example, the account type, debit and credit signs, etc. are standardized according to the metadata dictionary, and the transaction time field is uniformly processed in the format of year, month, day, hour, minute and second. For digital types such as amount, two decimal places are uniformly retained.

[0064] Furthermore, the key account analysis module (2) includes:

[0065] A grouping statistics unit (21) is used to group and count the fund transaction data based on the local account and the peer account to form aggregated transaction data;

[0066] A relationship graph construction unit (22) is connected to the grouping statistics unit (21) and is used to use the local account and the peer account in the aggregated transaction data as nodes and the transaction relationship between the local account and the peer account as edges to construct and save a transaction relationship graph;

[0067] The key account identification unit (23) and the connection relationship graph construction unit (22) are used to calculate the support of the edges in the transaction relationship graph using the k-truss algorithm, screen the nodes and edges according to the support of the edges to construct a k-truss subgraph, calculate the key nodes from the k-truss subgraph as key accounts, and form a key account analysis result.

[0068] The grouping statistics unit performs grouping statistics based on local accounts and peer accounts, calculates transaction frequency and total transaction amount, and generates aggregated transaction data.

[0069] The key account identification unit reads the transaction relationship graph consisting of the saved nodes and edges, and uses the k-truss algorithm to calculate the support of all edges in the transaction relationship graph. The support of an edge refers to the number of triangles that the edge participates in. The edges and nodes are screened according to the support of the edge to construct a k-truss subgraph.

[0070] It includes: gradually deleting nodes and their connected edges whose support is less than the support threshold -2 in the network, retaining the subgraph of nodes whose support is not less than the support threshold -2, and constructing a k-truss subgraph. For example, if the support threshold k is 3, then k-2 is equal to 1, indicating that the support of the edge is 1, indicating that the edge participates in at least forming 1 triangle with other edges. If k is 4, the support of the edge is 2, indicating that the edge participates in at least forming 2 triangles with other edges.

[0071] Afterwards, for each node in the k-truss subgraph, the degree or other importance indicators of the node in the k-truss subgraph (such as betweenness centrality, closeness centrality, etc.) are calculated to measure its importance in the entire network. The key node finally calculated is the key account.

[0072] The k-truss subgraph can be seen as a more densely connected subnetwork, which reflects the high connectivity of local areas in the network. Calculating the degree of a node in the k-truss subgraph is to measure the centrality of the node by the degree of the node. The degree of a node refers to the number of edges connected to the node. An important node is a node with many connections. Betweenness Centrality measures the degree to which a node acts as an intermediary for the shortest path between other nodes. Closeness Centrality refers to the degree of closeness between a node and other nodes. If the shortest distance from a node to other nodes is small, then the closeness centrality of this node is very good.

[0073] Key nodes as key accounts can be identified from the k-truss subgraph through one of the node's degree centrality, betweenness centrality, and closeness centrality.

[0074] There are two ways to identify key nodes through the degree centrality of nodes. One is to sort the nodes from large to small according to their degree, and intercept a predetermined number of nodes in the front as key nodes. The second way is to take nodes whose degree is greater than a predetermined degree threshold as key nodes.

[0075] Furthermore, the visualization display module (3) includes a relationship graph query unit (31): used for a user to input a query condition, the query condition including a support threshold;

[0076] The key account identification unit (23) is used to:

[0077] The transaction relationship subgraph is initially filtered according to the query conditions input by the user to form a k-truss subgraph. The initial filtering includes filtering out the edges and nodes whose edge support is greater than the support threshold to form a k-truss subgraph;

[0078] Calculate the key nodes from the k-truss subgraph as key accounts to form key account analysis results;

[0079] Return the key account analysis results shown to the visualization display module;

[0080] The visualization display module (3) also includes a visualization interaction unit (32), which is connected to the relationship graph query unit and is used to display the searched key account analysis results.

[0081] The system gives a support threshold by default, and users can enter a support threshold by themselves. The nodes and edges that meet the query conditions are obtained through the relationship graph query statement, and are encapsulated into the format required by the front-end component and returned.

[0082] In the visualization interaction unit, users can customize the style, size, color, etc. of key nodes and key accounts according to actual needs, as well as the style, thickness, color, etc. of edges. The visualization interaction unit supports user interaction operations, such as clicking on a node to expand detailed information, dragging a node to change the relationship layout, etc. This facilitates users to explore and understand complex network structures.

[0083] Preferably, the front end of the visualization interaction unit uses the G6 graph relationship network component in AntV visualization analysis, and provides a variety of layout algorithms, such as force-directed layout, circular layout, etc. Users can choose the appropriate layout algorithm according to their needs to best display the relationship between nodes and edges in the key account analysis results, and can handle large-scale nodes and edges to maintain a smooth interactive experience.

[0084] Furthermore, the attributes of the nodes in the transaction relationship graph include account number and account holder name;

[0085] The attributes of the edges in the transaction relationship graph include transaction frequency and transaction amount;

[0086] The query condition input by the user also includes at least one of transaction amount, transaction frequency, account number, and account holder name.

[0087] As a query condition, the user can also enter at least one of the transaction amount, transaction frequency, account number, and account holder name to identify targeted key accounts. These query conditions are initially filtered, and then further preliminary filtered according to the support degree of the edge.

[0088] See also Figure 2 The present invention provides a key account analysis method based on capital transactions, using the aforementioned key account analysis system based on capital transactions, including:

[0089] Step A1, accessing fund transaction data and preprocessing;

[0090] Step A2, using the k-truss algorithm to perform key account analysis on the pre-processed fund transaction data, automatically marking key accounts, and obtaining key account analysis results;

[0091] Step A3: Display the key account analysis results.

[0092] The present invention processes large-scale fund transaction data, constructs a transaction relationship graph, and identifies key accounts in the relationship graph based on the k-truss algorithm, with better recognition efficiency, accuracy, and real-time performance. Moreover, the overall key account can be identified, rather than just limited to a single account or multiple accounts, so as to adapt to different analysis needs and help improve the ability of financial institutions in fund flow monitoring and risk management.

[0093] See also Figure 3 , further, step A1 comprises:

[0094] Step A11, accessing fund transaction data;

[0095] Step A12, cleaning the fund transaction data according to preset cleaning rules;

[0096] Step A13, converting the cleaned fund transaction data according to preset data standardization rules to obtain standardized fund transaction data.

[0097] It supports real-time or batch data connection with other information systems through API interface or database connection, and supports the import of transaction data files in multiple formats (such as CSV, Excel, JSON, etc.) to process historical transaction records.

[0098] Data cleaning removes duplicate records from the accessed fund transaction data to ensure the uniqueness and accuracy of the data set. Estimate missing values ​​through account information or interpolation methods. Delete or modify or replace outliers based on the context.

[0099] The conversion normalizes the data for subsequent analysis and processing. For example, the account type, loan sign, etc. are standardized according to the metadata dictionary, the transaction time field is uniformly processed in the format of year, month, day, hour, minute, and second, and the numeric types such as the amount are uniformly processed with two decimal places.

[0100] See also Figure 4 , further, step A2 comprises:

[0101] Step A21, grouping and counting the fund transaction data based on the local account and the peer account to form aggregated transaction data;

[0102] Step A22, using the local account and the peer account in the aggregated transaction data as nodes, and the transaction relationship between the local account and the peer account as an edge, to construct and save a transaction relationship graph;

[0103] Step A23, use the k-truss algorithm to calculate the support of the edges in the transaction relationship graph, screen the nodes and edges according to the support of the edges to construct a k-truss subgraph, calculate the key nodes from the k-truss subgraph as key accounts, and form the key account analysis results.

[0104] The nodes whose support is less than the support threshold -2 and the edges connected to them in the relationship graph are gradually deleted, and the subgraph of nodes whose support is not less than the support threshold -2 is retained to construct the k-truss subgraph.

[0105] The k-truss subgraph can be regarded as a more densely connected subnetwork, which reflects the high connectivity of local areas in the network. For each node in the k-truss subgraph, the degree or other importance indicators (such as betweenness centrality, closeness centrality, etc.) of the node in the k-truss subgraph are calculated to measure its importance in the entire network. The key nodes finally calculated are the key accounts.

[0106] Furthermore, before step A23, step B1 is also included: the user inputs a query condition, and the query condition includes a support threshold;

[0107] In step A23, the transaction relationship subgraph is initially filtered according to the query conditions input by the user to form a k-truss subgraph, and the key nodes are calculated from the k-truss subgraph as key accounts to form and return the key account analysis results. The initial filtering includes screening out the edges and nodes whose edge support is greater than the support threshold.

[0108] Further, in step A22, the attributes of the nodes in the transaction relationship graph include the account number and the name of the account holder, and the attributes of the edges in the transaction relationship graph include the transaction frequency and the transaction amount;

[0109] In step B1, the query condition input by the user also includes at least one of transaction amount, transaction frequency, account number, and account holder name.

[0110] As a query condition, the user can also enter at least one of the transaction amount, transaction frequency, account number, and account holder name to identify targeted key accounts. These query conditions are initially filtered, and then further preliminary filtered according to the support degree of the edge.

[0111] The core principle of the k-Truss algorithm is to identify subgraphs with highly interconnected internal nodes in the transaction relationship graph. Its definition and detection process is based on the support of the edge. Specifically, each edge participates in at least k-2 triangles. The K-Truss algorithm can effectively extract highly cohesive subgraphs from large and complex networks, which is very valuable for understanding network structure characteristics and mining hidden information.

[0112] Key account analysis of capital transaction network. This invention regards each account as a node and each transaction as an edge to construct an undirected or directed graph. By calculating the support of the edge and the closeness between the node and its surrounding neighbors, the K-Truss algorithm can identify highly interconnected subgraphs in the graph. It can find possible core groups in a complex transaction network, thereby finding key account information in the capital network transaction relationship graph.

[0113] With the development of social economy, people and enterprises are increasingly engaging in capital transactions in their daily work and life. When analyzing capital transactions of certain suspicious entities, the amount of data that needs to be processed is generally large. The definition of k-Truss may seem complicated, but there are efficient algorithms to implement it. Especially for sparse graphs, some optimization strategies such as support threshold screening can be used to reduce the amount of calculation, which is highly efficient.

[0114] The analysis activity of the capital transaction network is often an ongoing one. In the process of analyzing a data set, newly acquired capital transaction data may be gradually added over time, which will cause the graph structure of the capital transaction network to change continuously. When new capital transaction data appears, the data access processing module accesses the newly added capital transaction data in an incremental manner. The grouping statistics unit processes the newly added capital transaction data and calculates the aggregated transaction data associated with the newly added capital transaction data in combination with the saved transaction relationship graph. The relationship graph construction unit incrementally updates the transaction relationship graph according to the aggregated transaction data associated with the newly added capital transaction data. The existing key account identification calculates the key account for the full amount of capital transaction data each time. The k-Truss dynamic update mechanism of the present invention provides better adaptability. When a new edge is added to the transaction relationship graph or an existing edge is removed, the entire transaction relationship graph does not need to be recalculated. Instead, only the affected part is locally updated. Similarly, when an edge is removed, a similar incremental method can be used to update the k-truss subgraph under the condition that the query condition remains unchanged, instead of calculating it all again. Financial institutions handle massive amounts of capital flows every day, and transaction networks change rapidly. The efficient dynamic adaptability of K-Truss can help track capital links in real time, detect abnormal activities in a timely manner, and enhance risk management capabilities. In general, the present invention updates the transaction relationship graph and k-truss subgraph in real time through an incremental update method, reducing a large number of repeated calculations. The present invention not only has high computational efficiency, but also saves computing resources.

[0115] In addition, when analyzing key accounts in a fund transaction network, users expect to obtain accurate calculation results for core accounts. If there are too many key node accounts that have been calculated and fed back, it is not convenient for further investigation and analysis. The present invention uses the K-Truss algorithm to find subgraphs whose internal nodes are highly interconnected. The nodes in these subgraphs are not only directly connected to each other, but also indirectly connected through multiple other nodes. This feature makes k-Truss an ideal choice for identifying closely related key accounts.

[0116] In addition, since k-Truss can be calculated for different support thresholds k, it can provide information about different levels of the transaction relationship graph. Lower-level k-Truss (i.e., smaller support thresholds) can obtain a wider range of key accounts, while higher-level k-Truss (i.e., larger support thresholds) focus on a tighter core. Financial institutions can track the core paths of capital flows through multi-level k-Truss analysis, detect abnormal activities in a timely manner, and enhance risk early warning capabilities. By gradually parsing the degree of nodes from low to high, the increasingly tight subgroups or communities in the network are revealed layer by layer, which helps to discover key accounts and connections hidden in the deep layers of the network but crucial to the overall structure, and observe which nodes are gradually excluded from the k-Truss subgraph and which nodes are always retained in the k-Truss subgraph with higher density, thus providing a basis for the final determination of key accounts.

[0117] The above are only preferred embodiments of the present invention, and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the description and illustrations of the present invention should be included in the protection scope of the present invention.

Claims

1. A key account analysis system based on capital transactions, characterized in that: include: Data collection and processing module, used for access and preprocessing of fund transaction data; A key account analysis module, connected to the data acquisition and processing module, is used to perform key account analysis on the pre-processed fund transaction data using a k-truss algorithm, automatically mark key accounts, and obtain key account analysis results; A visualization display module is connected to the key account analysis module and is used to display the key account analysis results.

2. A key account analysis system based on fund transactions as claimed in claim 1, characterized in that: The data acquisition and processing module comprises: A data access unit, used for accessing the fund transaction data; A data cleaning unit, connected to the data access unit, for cleaning the fund transaction data according to a preset cleaning rule; The data conversion unit is connected to the data cleaning unit and is used to convert the cleaned fund transaction data according to preset data standardization rules to obtain the standardized fund transaction data.

3. A key account analysis system based on fund transactions as claimed in claim 1, characterized in that: The key account analysis module includes: A grouping statistics unit, used for grouping and counting the fund transaction data based on the local account and the peer account to form aggregated transaction data; a relationship graph construction unit, connected to the group statistics unit, and configured to use the local account and the peer account in the aggregated transaction data as nodes, and the transaction relationship between the local account and the peer account as edges, to construct and save a transaction relationship graph; A key account identification unit is connected to the relationship graph construction unit and is used to calculate the support of the edges in the transaction relationship graph using the k-truss algorithm, screen nodes and edges according to the support of the edges to construct a k-truss subgraph, calculate the key nodes from the k-truss subgraph as the key accounts, and form the key account analysis results.

4. A key account analysis system based on fund transactions as claimed in claim 3, characterized in that: The visualization display module includes a relationship graph query unit: used for a user to input a query condition, wherein the query condition includes a support threshold; The key account identification unit is used to: Preliminarily filtering the transaction relationship subgraph according to the query condition input by the user to form the k-truss subgraph, wherein the preliminary filtering includes screening out edges and nodes whose edge support is greater than the support threshold corresponding to form the k-truss subgraph; Calculate key nodes from the k-truss subgraph as the key accounts to form the key account analysis results; Returning the key account analysis results shown to the visualization display module; The visualization display module also includes a visualization interaction unit, which is connected to the relationship graph query unit and is used to display the searched key account analysis results.

5. A key account analysis system based on fund transactions as claimed in claim 4, characterized in that: The attributes of the nodes in the transaction relationship graph include account number and account holder name; The attributes of the edge in the transaction relationship graph include transaction frequency and transaction amount; The query condition input by the user also includes at least one of transaction amount, transaction frequency, account number, and account holder name.

6. A key account analysis method based on capital transactions, characterized in that: A key account analysis system based on fund transactions as described in any one of claims 1 to 5 is used, comprising: Step A1, accessing fund transaction data and preprocessing; Step A2, using the k-truss algorithm to perform key account analysis on the pre-processed fund transaction data, automatically marking key accounts, and obtaining key account analysis results; Step A3: display the key account analysis results.

7. A key account analysis method based on fund transactions as claimed in claim 6, characterized in that: The step A1 comprises: Step A11, accessing the fund transaction data; Step A12, performing data cleaning on the fund transaction data according to a preset cleaning rule; Step A13, converting the cleaned fund transaction data according to preset data standardization rules to obtain standardized fund transaction data.

8. A key account analysis method based on fund transactions as claimed in claim 6, characterized in that: The step A2 comprises: Step A21, grouping and counting the fund transaction data based on the local account and the peer account to form aggregated transaction data; Step A22, using the local account and the peer account in the aggregated transaction data as nodes, and the transaction relationship between the local account and the peer account as an edge, to construct and save a transaction relationship graph; Step A23, using the k-truss algorithm to calculate the support of the edges in the transaction relationship graph, screening the nodes and edges according to the support of the edges to construct a k-truss subgraph, calculating the key nodes from the k-truss subgraph as the key accounts, and forming the key account analysis results.

9. A key account analysis method based on fund transactions as claimed in claim 8, characterized in that: The step A23 also includes step B1: the user inputs a query condition, and the query condition includes a support threshold; In the step A23, the transaction relationship subgraph is preliminarily filtered according to the query condition input by the user to form the k-truss subgraph, and the key nodes are calculated from the k-truss subgraph as the key accounts to form the key account analysis results and return them. The preliminary filtering includes screening out the edges and nodes whose edge support is greater than the support threshold.

10. A key account analysis method based on fund transactions as claimed in claim 9, characterized in that: In the step A22, the attributes of the nodes in the transaction relationship graph include the account number and the name of the account holder, and the attributes of the edges in the transaction relationship graph include the transaction frequency and the transaction amount; In step B1, the query condition input by the user also includes at least one of transaction amount, transaction frequency, account number, and account holder name.

Images