Data mining method and device, computer device and storage medium

Abstract

The application relates to a data mining method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring object resource data; constructing an object feature graph according to the object resource data; acquiring each abnormal node in the object feature graph; respectively calculating influence factors corresponding to each abnormal node and each other node in the object feature graph; determining each target node in the other nodes according to the influence factors corresponding to each other node, and generating corresponding correlation factors according to the influence factors of each target node; and generating target information correlation data based on the correlation paths between the abnormal nodes and the correlation factors, wherein the target information correlation data is used to represent the correlation between each target node and each abnormal node. The method can effectively improve the accuracy of data mining.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a data mining method, apparatus, computer device, and storage medium. Background Technology

[0002] With the development of computer technology, automation has become increasingly widespread in various industries, which has led to a huge demand for information mining of large amounts of data. Among these, graph data mining has gradually become an important branch of the data mining field.

[0003] Most existing technologies employ visual clustering analysis to locally cluster nodes and connections in various regions of graph data, but the accuracy of data mining is poor. Summary of the Invention

[0004] Therefore, it is necessary to provide a data mining method, apparatus, computer equipment, and computer-readable storage medium to address the aforementioned technical problems, which can effectively improve the accuracy of data mining.

[0005] A data mining method, characterized in that the method includes:

[0006] Retrieve object resource data;

[0007] An object feature map is constructed based on the object resource data. The object feature map is used to characterize the node attributes of each node in the object resource data and the association paths between each node.

[0008] Obtain each abnormal node in the object feature map;

[0009] Calculate the influence factors corresponding to each abnormal node and each other node in the object feature map;

[0010] Based on the influence factors corresponding to each other node, determine each target node among the other nodes, and generate corresponding association factors based on the influence factors of each target node. The association factors are used to characterize the degree of association between the target node and the corresponding abnormal node under the corresponding association path.

[0011] Target information association data is generated by fusing the association paths and association factors between each abnormal node. This target information association data is used to characterize the association relationship between each target node and each abnormal node.

[0012] In one embodiment, the influence factors corresponding to each abnormal node and each other node in the object feature map are calculated, including:

[0013] Obtain the associated paths between each abnormal node and each other node;

[0014] Calculate the impact factors of abnormal nodes and other nodes on each associated path.

[0015] In one embodiment, each target node among the other nodes is determined based on the influence factors corresponding to each other node, and a corresponding correlation factor is generated based on the influence factors of each target node, including:

[0016] When the impact factor is greater than the preset threshold, other nodes corresponding to the impact factor are identified as target nodes.

[0017] Obtain the previous time step association factor and propagation factor of the target node. The propagation factor is used to characterize the transition probability of the corresponding influence factor between adjacent nodes.

[0018] The corresponding correlation factor is obtained by fusing the correlation factor, propagation factor, and corresponding influence factor from the previous time step.

[0019] In one embodiment, after determining each target node among the other nodes based on the influence factors corresponding to each other node, and generating corresponding correlation factors based on the influence factors of each target node, the method further includes:

[0020] Identify the neighboring nodes of the target node and obtain the adjacent association paths between the target node and its neighboring nodes;

[0021] The fusion association path between the abnormal node and its adjacent nodes is obtained based on the adjacent association paths and the association paths between the abnormal node and the target node.

[0022] Obtain the previous time-influence factors of the abnormal node and its adjacent nodes;

[0023] The influence factors of the abnormal node and the target node are fused with the influence factors of the previous moment to obtain the influence factors of the abnormal node and its neighboring nodes at the current moment.

[0024] Return to the steps for calculating the influence factors corresponding to each anomaly node and each other node in the object feature map.

[0025] In one embodiment, obtaining the fused association path between the abnormal node and its neighboring nodes based on the adjacent association paths and the association path between the abnormal node and the target node includes:

[0026] The target node on the path associated with the abnormal node and the target node is taken as the termination node.

[0027] The terminating node is used as the starting point of adjacent associated paths to splice the associated paths and generate a merged associated path.

[0028] In one embodiment, target information association data is generated by fusing the association paths and association factors between various abnormal nodes, including:

[0029] Obtain the set of association factors corresponding to all associated paths between each abnormal node and each target node;

[0030] The target information association data is obtained by summing the association paths between the set of association factors and each abnormal node.

[0031] In one embodiment, after generating target information association data based on the fusion of association paths and association factors between various abnormal nodes, the method further includes:

[0032] When the amount of data associated with the target information exceeds a preset threshold, the target node in the associated path corresponding to the data associated with the target information is determined to be in an abnormal state.

[0033] When the target information associated data is less than or equal to a preset threshold, the target node of the associated path corresponding to the target information associated data is determined to be in a normal state.

[0034] A data mining apparatus, characterized in that the apparatus comprises:

[0035] The data processing module is used to acquire object resource data; construct an object feature map based on the object resource data, which is used to characterize the node attributes of each node in the object resource data and the association paths between each node;

[0036] The impact factor generation module is used to obtain each anomalous node in the object feature map; and to calculate the impact factor corresponding to each anomalous node and each other node in the object feature map.

[0037] The correlation factor generation module is used to determine each target node among other nodes based on the influence factors corresponding to each other node, and to generate corresponding correlation factors based on the influence factors of each target node. The correlation factors are used to characterize the degree of correlation between the target node and the corresponding abnormal node under the corresponding correlation path.

[0038] The target information association data generation module is used to generate target information association data based on the association paths and association factors between various abnormal nodes. The target information association data is used to characterize the association relationship between each target node and each abnormal node.

[0039] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0040] Retrieve object resource data;

[0041] An object feature map is constructed based on the object resource data. The object feature map is used to characterize the node attributes of each node in the object resource data and the association paths between each node.

[0042] Obtain each abnormal node in the object feature map;

[0043] Calculate the influence factors corresponding to each abnormal node and each other node in the object feature map;

[0044] Based on the influence factors corresponding to each other node, determine each target node among the other nodes, and generate corresponding association factors based on the influence factors of each target node. The association factors are used to characterize the degree of association between the target node and the corresponding abnormal node under the corresponding association path.

[0045] Target information association data is generated by fusing the association paths and association factors between each abnormal node. This target information association data is used to characterize the association relationship between each target node and each abnormal node.

[0046] A computer-readable storage medium storing a computer program, which, when executed by a processor, performs the following steps:

[0047] Retrieve object resource data;

[0048] An object feature map is constructed based on the object resource data. The object feature map is used to characterize the node attributes of each node in the object resource data and the association paths between each node.

[0049] Obtain each abnormal node in the object feature map;

[0050] Calculate the influence factors corresponding to each abnormal node and each other node in the object feature map;

[0051] Based on the influence factors corresponding to each other node, determine each target node among the other nodes, and generate corresponding association factors based on the influence factors of each target node. The association factors are used to characterize the degree of association between the target node and the corresponding abnormal node under the corresponding association path.

[0052] Target information association data is generated by fusing the association paths and association factors between each abnormal node. This target information association data is used to characterize the association relationship between each target node and each abnormal node.

[0053] The aforementioned data mining method, apparatus, computer equipment, and storage medium acquire object resource data, construct an object feature map based on the object resource data, then acquire each abnormal node in the object feature map, calculate the influence factors corresponding to each abnormal node and each other node in the object feature map, determine each target node among the other nodes based on the influence factors corresponding to each other node, and generate corresponding correlation factors based on the influence factors of each target node. Finally, target information correlation data is generated by fusing the correlation paths and correlation factors between each abnormal node. In this way, by iteratively calculating the influence factors and correlation factors of each other node and abnormal node in the graph data regarding the corresponding correlation paths through the topological structure and node attributes of each node, the internal features between each node attribute and correlation path are effectively combined, fully mining the correlation relationships between nodes under corresponding paths, and effectively improving the accuracy of data mining. Attached Figure Description

[0054] Figure 1 This is a diagram illustrating the application environment of a data mining method in one embodiment;

[0055] Figure 2 This is a flowchart illustrating a data mining method in one embodiment;

[0056] Figure 3 This is a schematic diagram of the process for generating impact factors in one embodiment;

[0057] Figure 4 This is a schematic diagram of the process for generating correlation factors in one embodiment;

[0058] Figure 5 This is a schematic diagram of the process for updating the current impact factor in one embodiment;

[0059] Figure 6 This is a flowchart illustrating the process of generating fusion association paths in one embodiment;

[0060] Figure 7 This is a schematic diagram of the process for generating target information association data in one embodiment;

[0061] Figure 8 This is a flowchart illustrating the process of determining the state of a target node in one embodiment;

[0062] Figure 9 This is a structural block diagram of a data mining apparatus in one embodiment;

[0063] Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0065] The data mining method provided in this application embodiment can be applied to, for example... Figure 1 The application environment shown. For example... Figure 1 As shown, computer device 102 acquires object resource data, constructs an object feature map based on the object resource data, acquires each abnormal node in the object feature map, calculates the influence factor corresponding to each abnormal node and each other node in the object feature map, determines each target node among the other nodes based on the influence factors corresponding to each other node, and generates corresponding association factors based on the influence factors of each target node. Finally, it generates target information association data by fusing the association paths and association factors between each abnormal node. Specifically, computer device 102 may include, but is not limited to, various personal computers, laptops, servers, smartphones, tablets, smart cameras, and portable wearable devices.

[0066] In one embodiment, such as Figure 2 As shown, a data mining method is provided, which can be applied to... Figure 1 Taking computer device 102 as an example, the following steps are included:

[0067] Step S202: Obtain object resource data.

[0068] Object resource data is used to represent object data with multiple dimensions. For example, object resource data can be book resource data, including dimensions such as book publication time, book category, book author's nationality, book publisher, and book word count range. It can also be user resource data, including dimensions such as user contact information, bank card, and enterprise.

[0069] Step S204: Construct an object feature map based on the object resource data. The object feature map is used to characterize the node attributes of each node in the object resource data and the association paths between each node.

[0070] Specifically, the computer device constructs an object feature graph based on the node data (i.e., each dimension) and the association paths between the nodes in the object resource data obtained in the aforementioned steps. That is, the nodes in the object feature graph are used to represent the node data of each dimension in the object resource data, and the edges in the object feature graph (i.e., the edges between each node) are used to represent the association paths between the nodes of each dimension in the object resource data.

[0071] Step S206: Obtain each abnormal node in the object feature map.

[0072] Specifically, the computer device defines the dimension nodes with preset characteristics in the object resource data as target nodes or abnormal nodes, and then determines the position of these abnormal nodes in the object feature map based on the object feature map, and obtains the position information of these abnormal nodes in the object feature map and all associated paths with other nodes, wherein there can be one or more abnormal nodes.

[0073] Step S208: Calculate the influence factors corresponding to each abnormal node and each other node in the object feature map.

[0074] The influence factor is used to characterize the error between the estimated importance value and the true importance value of the abnormal node and other nodes in the object feature map. The larger the influence factor, the closer the estimated importance value of the corresponding abnormal node and other nodes is to the true importance value.

[0075] Specifically, the computer equipment calculates the importance residuals of the influence factors corresponding to each abnormal node and each other node in the object feature map, and then uses the importance residuals as influence factors. The importance residuals are used to characterize the residuals between the importance estimates of the abnormal node and the corresponding other nodes. The larger the residual, the greater the error in the correlation estimate between the current abnormal node and the corresponding other factors.

[0076] Step S210: Determine each target node among the other nodes based on the influence factors corresponding to each other node, and generate corresponding association factors based on the influence factors of each target node. The association factors are used to characterize the degree of association between the target node and the corresponding abnormal node under the corresponding association path.

[0077] Specifically, the computer device determines the influence factors of each abnormal node and other nodes according to the aforementioned steps, and then compares the magnitude of each influence factor with a preset threshold. When the influence factor is less than the preset threshold, it indicates that the error between the estimated importance value and the true importance value between the current node and the corresponding abnormal node is greater than the preset threshold. This indicates that the current node needs to undergo subsequent iterations. Therefore, the computer selects this current node as the target node, and so on to determine all target nodes in the entire object feature map. Then, it obtains the previous moment's importance estimate between the current node and the corresponding other nodes, and fuses the previous moment's importance estimate with the influence factor to obtain the correlation factor (i.e., the current moment's importance estimate). The correlation factor is used to characterize the degree of correlation between the target node and the corresponding abnormal node under the corresponding correlation path. The larger the value of the correlation factor, the greater the correlation between the current abnormal node and the corresponding other nodes.

[0078] Step S212: Based on the fusion of the association paths and association factors between each abnormal node, target information association data is generated. The target information association data is used to characterize the association relationship between each target node and each abnormal node.

[0079] Specifically, the computer equipment determines the correlation factors corresponding to each abnormal node and each other node in the object feature map according to the aforementioned steps, and then weights and fuses the correlation factors and the correlation paths between each abnormal node to obtain target information correlation data. The information correlation data is used to characterize the correlation relationship between each target node and each abnormal node.

[0080] In this embodiment, object resource data is acquired, an object feature graph is constructed based on the object resource data, and then each abnormal node in the object feature graph is acquired. The influence factors corresponding to each abnormal node and each other node in the object feature graph are calculated. Based on the influence factors corresponding to each other node, each target node in the other nodes is determined, and the corresponding association factors are generated based on the influence factors of each target node. Target information association data is generated by fusing the association paths and association factors between each abnormal node. In this way, by iteratively calculating the influence factors and association factors of each other node and abnormal node in the graph with respect to the corresponding association paths through the topological structure of each node in the graph data and the node attributes of each node, the internal features between each node attribute and the association path are effectively combined, fully exploring the association relationships between nodes under the corresponding paths, and effectively improving the accuracy of data mining.

[0081] In one embodiment, such as Figure 3 As shown, the influence factors corresponding to each abnormal node and each other node in the object feature map are calculated respectively, including:

[0082] Step S302: Obtain the associated paths between each abnormal node and each other node.

[0083] Among them, the various association paths between each abnormal node and each other node refer to the connection paths between each abnormal node and each other node in the object feature map.

[0084] Specifically, the computer device acquires the location information of each node in the object feature map and the connection information between each node. Based on this information, it constructs a node and path database, and then obtains the associated paths between each abnormal node and each other node.

[0085] Step S304: Calculate the influence factors of the abnormal node and each other node on each associated path.

[0086] Each of these influencing factors pertains to a specific association path, and the influencing factors for the same two nodes can differ under different association paths.

[0087] Specifically, after the computer equipment determines the paths between each node in the object feature map according to the aforementioned steps, it then calculates the influence factor corresponding to each node under its respective path.

[0088] In this embodiment, the computer device traverses all the associated paths between nodes by examining the positional relationships and connections between nodes in the object feature graph. Then, it traverses the influence factors of abnormal nodes and other nodes in the entire object feature graph under various associated path conditions. This fully explores the attributes between nodes and the features on the paths in the object feature graph, and calculates the influence factors between nodes based on the features on the path associations, effectively improving the accuracy of the influence factor calculation.

[0089] In one embodiment, such as Figure 4 As shown, each target node among the other nodes is determined based on its corresponding influence factor, and corresponding correlation factors are generated based on the influence factors of each target node, including:

[0090] Step S402: When the influence factor is greater than a preset threshold, other nodes corresponding to the influence factor are determined as target nodes.

[0091] Specifically, the computer device obtains the preset threshold corresponding to the impact factor, and then compares the impact factor of each point determined in the previous steps with the preset threshold. When the impact factor is greater than the preset threshold, it indicates that the error between the estimated importance value of the current abnormal node and the corresponding other nodes in the current path and the true importance value is too large. Therefore, it is necessary to perform subsequent iterative steps on the estimated importance value of the current abnormal node and the current other nodes in the current path to obtain an importance estimate value that is closer to the true importance value.

[0092] Step S404: Obtain the previous time-series association factor and propagation factor of the target node. The propagation factor is used to characterize the transition probability of the corresponding influence factor between adjacent nodes.

[0093] Step S406: Based on the previous time step, the correlation factor, propagation factor and corresponding influence factor are fused to obtain the corresponding correlation factor.

[0094] Specifically, the computer device obtains the correlation factor, propagation factor, and influence factor of the target node at the previous moment, and obtains the correlation factor iteration term by fusing the propagation factor and the influence factor at the current moment. Then, it obtains the corresponding correlation factor (i.e., the correlation factor at the current moment) by combining the correlation factor iteration term with the correlation factor at the previous moment.

[0095] For example, computer devices can construct the corresponding correlation factors according to the method shown in Formula 1 below:

[0096]

[0097] in, The correlation factor at the current moment, Let r′ be the correlation factor from the previous time step, α be the propagation factor, and r′ be the propagation factor. s (u)(p) represents the influence factor at the current moment.

[0098] In this embodiment, the computer device obtains the previous correlation factor and propagation factor of the target node, and then obtains the corresponding correlation factor by fusing the previous correlation factor, propagation factor and corresponding influence factor. By setting the propagation factor in the iterative formula, the correlation factor at the current moment is updated according to the current influence factor with a certain propagation ratio, thereby improving the accuracy of the correlation factor iteration.

[0099] In one embodiment, such as Figure 5 As shown, after determining each target node among the other nodes based on the influence factors corresponding to each other node, and generating corresponding correlation factors based on the influence factors of each target node, the process also includes:

[0100] Step S502: Determine the neighboring nodes of the target node and obtain the adjacent association paths between the target node and its neighboring nodes.

[0101] Specifically, the computer device acquires the location information of the target node in the object feature map, and then, with the target node as the center, acquires the multiple connections between the target node and other nodes, thereby determining all the adjacent nodes of the target node and all the adjacent associated paths between the target node and its adjacent nodes.

[0102] Step S504: Obtain the fusion association path between the abnormal node and its neighboring nodes based on the adjacent association paths and the association path between the abnormal node and the target node.

[0103] Specifically, the computer device obtains all adjacent associated paths of the target node according to the aforementioned steps, then obtains the associated path between the target node and the abnormal node, and then merges this associated path with the adjacent associated paths to obtain the merged associated path.

[0104] For example, computer devices can merge associated paths according to the method shown in Formula 2 below, where node u and node v are connected through nodes x1, x2, ..., x... m The associated path pattern is represented as follows:

[0105] p1+p2=(φ V (u),φE ((u,x1)),φ V (x1),φ E ((u,x2)),φ V (x2),…,φ V (v)), Formula 2

[0106] Where, p1=(φ V (u),φ E ((u,x1)),…,φ V (x k )) represents path 1; p2 = (φ V (x k ),φ E ((x k ,x k+1 )),…,φ V (v)) represents associated path 2, φ V φ represents the mapping relationship between nodes and node types. E This represents the mapping relationship between the edges and the edge types.

[0107] Step S506: Obtain the previous time-influence factors corresponding to the abnormal node and its neighboring nodes.

[0108] Specifically, the computer equipment obtains the influence factors of the abnormal node and its neighboring nodes at the previous time step under the corresponding associated paths.

[0109] Step S508: The influence factors of the abnormal node and the target node are fused with the influence factors of the previous moment to obtain the influence factors of the abnormal node and the adjacent nodes at the current moment.

[0110] Specifically, the computer device can update the current-time influence factors of the abnormal node and its neighboring nodes in the manner shown in Formula 3 below:

[0111]

[0112] Where, Δp=(φ V (u),φ E ((u,v)),φ V (v) represents the adjacent paths between the target node and its neighboring nodes, p represents the associated paths between the target node and the abnormal node, and p is the propagation factor. For the influence factor of the previous time step under the associated path p+Δp, Let r′ be the influence factor at the current time point under the associated path p+Δp. s (u)(p) represents the current impact factor under the associated path p.

[0113] Step S510 returns to the step of calculating the influence factors corresponding to each abnormal node and each other node in the object feature map.

[0114] Specifically, after the computer device updates the correlation factors of each abnormal node and each other node with respect to all paths according to the aforementioned steps to obtain the influence factors of all nodes at the current time, it returns to the step of calculating the influence factors corresponding to each abnormal node and each other node in the object feature map respectively.

[0115] In this embodiment, the computer device determines the neighboring nodes of the target node and obtains the adjacent association paths between the target node and the neighboring nodes. Based on the adjacent association paths and the association paths between the abnormal node and the target node, the fused association path between the abnormal node and the neighboring nodes is obtained. The previous moment's influence factor corresponding to the abnormal node and the neighboring node is obtained. Then, the influence factor between the abnormal node and the target node is fused with the previous moment's influence factor to obtain the current moment's influence factor between the abnormal node and the neighboring nodes. Finally, the steps of calculating the influence factors corresponding to each abnormal node and each other node in the object feature map are returned. This fully explores the attributes between each node in the object feature map and the features on the path. The influence factors between each node are calculated based on the features on the path association, which effectively improves the accuracy of the influence factor calculation.

[0116] In one embodiment, such as Figure 6 As shown, the fused association path between the abnormal node and its neighboring nodes is obtained based on the adjacent association paths and the association paths between the abnormal node and the target node, including:

[0117] Step S602: The target node on the associated path between the abnormal node and the target node is taken as the termination node.

[0118] Specifically, the computer device can determine the termination node in the manner shown in Formula 4 below:

[0119] Suppose that nodes u and v are connected by nodes x1, x2, ..., xn. m The associated path pattern is represented as follows:

[0120] p1=(φ V (u),φ E ((u,x1)),…,φ V (x k )), formula 4

[0121] Where p1 is the association path between the abnormal node and the target node, u is the abnormal node, and x k As the target node, the computer device will [x] k As the terminating node of this associated path.

[0122] Step S604: Use the termination node as the starting point of adjacent associated paths to splice the associated paths and generate a merged associated path.

[0123] Specifically, the computer equipment generates a merged and associated path by concatenating the paths as shown in Formula 5 below:

[0124] p1+p2=(φ V (u),φ E ((u,x1)),φ V (x1),φ E ((u,x2)),φ V (x2),…,φ V (v)), Formula 5

[0125] The computer device will terminate at point x in step S602. k As the associated path p2=(φ V (x k ),φ E ((x k ,x k+1 )),…,φ V (v)) Starting point, then concatenate p1 and p2 to generate the fusion associated path p1+p2.

[0126] In this embodiment, the computer device uses the target node on the association path between the abnormal node and the target node as the termination node, and then uses the termination node as the starting point of the adjacent association path to splice the association path to generate a fused association path, thereby realizing the splicing between two different paths in the object feature map and improving the efficiency of generating fused association paths.

[0127] In one embodiment, such as Figure 7 As shown, target information association data is generated by fusing the association paths and association factors between various abnormal nodes, including:

[0128] Step S702: Obtain the set of association factors corresponding to all associated paths between each abnormal node and each target node.

[0129] Step S704: Summing the association path between the association factor set and each abnormal node to obtain the target information association data.

[0130] Specifically, the computer device iterates over the association factors of all other nodes and each abnormal node in the entire object feature map with respect to their respective association paths according to the aforementioned steps until the iteration ends, thereby obtaining the set of association factors of each abnormal node with respect to all paths with respect to all paths. Then, it calculates the association path information between each abnormal node based on the object feature map, and finally fuses the association path information and association factors between each abnormal node to obtain the target information association data.

[0131] For example, a computer device can generate target information associated data PT(p) in the manner shown in Formula 6 below:

[0132] PT(p) = PT(p) + π s (u)(p), Formula 6

[0133] π s (u)(p) represents the set of correlation factors. This represents all overdue customer nodes. The path patterns and importance estimates between overdue customer node s and other overdue customer nodes are saved to PT.

[0134] In this embodiment, the computer device obtains the set of association factors corresponding to all association paths between each abnormal node and each target node, and sums the set of association factors with the association paths between each abnormal node to obtain target information association data. This fully explores the association relationships between abnormal nodes and other nodes, as well as between abnormal nodes, so that the generated target information association data can fully contain the semantic information contained in the object resource data, thereby improving the accuracy of data mining.

[0135] In one embodiment, such as Figure 8 As shown, after generating target information association data based on the fusion of association paths and association factors between various abnormal nodes, it also includes:

[0136] Step S802: When the target information associated data is greater than the preset threshold, the target node of the associated path corresponding to the target information associated data is determined to be in an abnormal state.

[0137] Specifically, the computer device compares the target information association data of the abnormal node and other corresponding nodes under the corresponding association path with a preset threshold. When the target information association data is greater than the preset threshold, it indicates that the degree of association between the target node corresponding to the current target information association data and the corresponding abnormal node is very high. Therefore, the computer device regards the current target as being in the same type as the abnormal node, and thus determines the current target node as an abnormal state.

[0138] Step S804: When the target information associated data is less than or equal to a preset threshold, the target node of the associated path corresponding to the target information associated data is determined to be in a normal state.

[0139] Specifically, the computer device compares the target information association data of the abnormal node and other corresponding nodes under the corresponding association path with a preset threshold. When the target information association data is less than or equal to the preset threshold, it indicates that the degree of association between the target node corresponding to the current target information association data and the corresponding abnormal node is very low. Therefore, the computer device regards the current target as not belonging to the same type as the abnormal node, and thus judges the current target node as normal.

[0140] In this embodiment, when the target information associated data is greater than a preset threshold, the target node of the associated path corresponding to the target information associated data is determined to be in an abnormal state. When the target information associated data is less than or equal to the preset threshold, the target node of the associated path corresponding to the target information associated data is determined to be in a normal state. The target node state determined based on the target information associated data generated in the aforementioned steps can more accurately reflect the state information of the target node and effectively improve the accuracy of data mining.

[0141] This application also provides an application scenario in which the above-mentioned data mining method is applied to predict the risk of loan delinquency in user accounts. Specifically, the data mining method is applied in this scenario as follows:

[0142] First, a heterogeneous graph is constructed based on real-world data and specific business operations. Nodes represent entities, including "customers," "contact information," "bank cards," and "enterprises," while edges represent relationships between entities, such as "employed at" and "linked bank cards." Then, a personalized heterogeneous path pattern perception algorithm is used to find the path patterns between a given overdue customer node and other overdue customer nodes. Finally, the critical path pattern is identified from the filtered path patterns.

[0143] Construct a heterogeneous graph. Based on the real world and specific business needs, consider representing the heterogeneous graph as G = (V, E, T). V ,T E ,φ V ,φ E ), where V represents the set of nodes, E represents the set of edges, and T V T represents the set of node types. E φ represents the set of edge types. V :V→T V φ represents the mapping relationship between nodes and node types. E :E→T EThis represents the mapping relationship between edges and edge types. For example, for nodes u∈V and v∈V, the edge between them is (u,v)∈E, and the types of nodes u and v are φ respectively. V (u)∈T V and φ V (v)∈T V The connection type between the two is φ E ((u,v))∈T E .

[0144] A personalized heterogeneous path pattern awareness algorithm is used to find the path patterns between a given overdue customer node and other overdue customer nodes, and to determine the criticality of each path pattern. First, a path pattern is defined, considering u,v∈V and m nodes x1,x2,…,x… m Node u and node v are connected through nodes x1, x2, ..., x m The associated path pattern is represented as follows:

[0145] (φ V (u),φ E ((u,x1)),φ V (x1),φ E ((x1,x2)),φ V (x2),…,φ E ((x m ,v)),φ V (v))∈P, Formula 7

[0146] Where P represents the set of various path patterns. Furthermore, for the following two paths p1∈P and p2∈P:

[0147] p1=(φ V (u),φ E ((u,x1)),…,φ V (x k )), formula 8

[0148] p2=(φ V (x k ),φ E ((x k ,x k+1 )),…,φ V (v)), Formula 9

[0149] Define the "path pattern sum", i.e., p1 + p2, as:

[0150] p1+p2=(φ V (u),φ E ((u,x1)),φ V (x1),φ E((u,x2)),φ V (x2),…,φ V (v)), Formula 10

[0151] Specifically, the empty path pattern is represented by θ, and

[0152] Referring to the way personalized webpage ranking defines importance, π represents the importance of other nodes in the heterogeneous graph to a given overdue customer node s, and this importance is propagated through a certain path pattern; s :V→[P→[0,1]] represents the importance estimate of other nodes in the heterogeneous graph to a given overdue node s, and this importance estimate is propagated through a certain path pattern; r s :V→[P→[0,1]] represents the importance residual of other nodes in the heterogeneous graph to a given overdue node s, and this importance residual is propagated through a certain path pattern. Specifically, the relationship between importance, importance estimate, and importance residual is as follows: (That is, the smaller the importance residual, the closer the importance estimate is to the precise value of importance); ∈ (0,1) represents the importance residual threshold, typically 10. -3 10 -4 10 -5 α∈(0,1) is the transition probability, representing the probability that the importance residual propagates from a node in the heterogeneous graph to its neighboring nodes, typically taking a value of 0.85. For example, for a node u∈V and a certain path pattern p∈P, π s (u)(p) indicates that a given overdue customer node s can be associated with node u through a path of pattern p, and the importance estimate of node u to the source node s is π under this path pattern. s (u)(p);r s (u)(p) indicates that the source node s can be associated with node u through a path of pattern p, and the importance residual of node u to the source node s is r under this path pattern. s (u)(p).

[0153] The following is the detailed process of the personalized heterogeneous path pattern awareness algorithm:

[0154] Step 1: Obtain the heterogeneous financial credit graph G = (V, E, T) consisting of entities such as "customer", "contact information", "bank card", and "enterprise", and relationships such as "employed at" and "linked bank card". V ,T E ,φ V ,φ E Identify a group of overdue customer nodes. The residual threshold ∈ and the transition probability α.

[0155] Step 2: Define an empty hash table variable PT, where PT(p) represents the criticality of path p.

[0156] Step 3: For V s For each overdue customer node s in the process, perform the following steps:

[0157] Step 3.1: Initialize variable π s and r s , And r s (s)(θ) = 1.0, meaning that the importance estimate of any node to node s through any path pattern is 0, the importance residual of any node other than node s to node s through any path pattern is 0, and the importance residual of node s to itself is 1.0.

[0158] Step 3.2: Iteratively calculate π s ,when ∑ p∈P r s When (u)(p)>:

[0159] Step 3.2.1: Transfer r s (u) Copy to r′ s (u) Then clear r s (u), r′ s (u)(p)=r s (u)(p)and r s (u)(p)=0,

[0160] Step 3.2.2: π s (u)(p)=π s (u)(p)+(1-α)r′ s (u)(p), Formula 11

[0161] Step 3.2.3: Let E(u) represent all edges associated with node u, then

[0162] Δp=(φ V (u),φ E ((u,v)),φ V (v)), Formula 12

[0163] r s (v)(p+Δp)=r s (v)(p+Δp)+αr′ s (u)(p), Formula 13

[0164] Step 3.3: Consider This represents all overdue customer nodes. The path patterns and importance estimates between overdue customer node s and other overdue customer nodes are saved to PT.

[0165]

[0166] Step 4: Output PT

[0167] Based on the output of the personalized heterogeneous path pattern perception algorithm in the second step, different path patterns are filtered out, and the critical path pattern is identified from them. Let the critical path pattern threshold be σ∈(0,1), typically σ=∈, then the set of critical path patterns is... for:

[0168]

[0169] The aforementioned data mining method acquires object resource data, constructs an object feature graph based on this data, then identifies each anomalous node within the feature graph, calculates the influence factors between each anomalous node and other nodes in the feature graph, determines target nodes based on these influence factors, and generates corresponding correlation factors based on the influence factors of each target node. Finally, it fuses the correlation paths and correlation factors between anomalous nodes to generate target information correlation data. In this way, by iteratively analyzing the topological structure and node attributes of each node in the graph data, the influence factors and correlation factors between other nodes and anomalous nodes regarding their corresponding correlation paths are derived. This effectively combines the internal features between node attributes and correlation paths, fully exploring the correlation relationships between nodes along corresponding paths and significantly improving the accuracy of data mining.

[0170] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0171] In one embodiment, such as Figure 9As shown, a data mining device is provided. This device can be a software module, a hardware module, or a combination of both integrated into a computer device. Specifically, the device includes: a data processing module 902, an influence factor generation module 904, a correlation factor generation module 906, and a target information correlation data generation module 908, wherein:

[0172] The data processing module 902 is used to acquire object resource data; construct an object feature map based on the object resource data, and the object feature map is used to characterize the node attributes of each node in the object resource data and the association paths between each node;

[0173] The impact factor generation module 904 is used to obtain each abnormal node in the object feature map; and to calculate the impact factor corresponding to each abnormal node and each other node in the object feature map respectively.

[0174] The correlation factor generation module 906 is used to determine each target node among other nodes based on the influence factors corresponding to each other node, and to generate corresponding correlation factors based on the influence factors of each target node. The correlation factors are used to characterize the degree of correlation between the target node and the corresponding abnormal node under the corresponding correlation path.

[0175] The target information association data generation module 908 is used to generate target information association data based on the fusion of association paths and association factors between various abnormal nodes. The target information association data is used to characterize the association relationship between each target node and each abnormal node.

[0176] The aforementioned data mining device acquires object resource data, constructs an object feature map based on the object resource data, then acquires each abnormal node in the object feature map, calculates the influence factors corresponding to each abnormal node and each other node in the object feature map, determines each target node among the other nodes based on the influence factors corresponding to each other node, and generates corresponding correlation factors based on the influence factors of each target node. Finally, it generates target information correlation data by fusing the correlation paths and correlation factors between each abnormal node. In this way, by iteratively calculating the influence factors and correlation factors of each other node and abnormal node in the graph data regarding the corresponding correlation paths through the topological structure and node attributes of each node, it effectively combines the internal features between each node attribute and the correlation path, fully explores the correlation relationships between nodes under the corresponding paths, and effectively improves the accuracy of data mining.

[0177] In one embodiment, the impact factor generation module 904 is further used to obtain the various association paths between each abnormal node and each other node; and to calculate the impact factor corresponding to each abnormal node and each other node on each association path.

[0178] In one embodiment, the correlation factor generation module 906 is further configured to determine other nodes corresponding to the influence factor as target nodes when the influence factor is greater than a preset threshold; obtain the previous correlation factor and propagation factor of the target node, wherein the propagation factor is used to characterize the transition probability of the corresponding influence factor between adjacent nodes; and obtain the corresponding correlation factor by fusing the previous correlation factor, propagation factor and corresponding influence factor.

[0179] In one embodiment, the association factor generation module 906 is further configured to determine the neighboring nodes of the target node and obtain the adjacent association paths between the target node and the neighboring nodes; obtain the fused association path between the abnormal node and the neighboring nodes based on the adjacent association paths and the association path between the abnormal node and the target node; obtain the previous time-influence factor corresponding to the abnormal node and the neighboring node; fuse the influence factor between the abnormal node and the target node with the influence factor at the previous time-influence factor to obtain the current time-influence factor between the abnormal node and the neighboring node; and return to the steps of calculating the influence factors corresponding to each abnormal node and each other node in the object feature map respectively.

[0180] In one embodiment, the association factor generation module 906 is further configured to use the target node on the association path between the abnormal node and the target node as the termination node; and to use the termination node as the starting point of the adjacent association path to splice the association path to generate a fused association path.

[0181] In one embodiment, the target information association data generation module 908 is further configured to obtain the set of association factors corresponding to all association paths between each abnormal node and each target node; and to sum the set of association factors and the association paths between each abnormal node to obtain the target information association data.

[0182] In one embodiment, the target information association data generation module 908 is further configured to determine the target node of the association path corresponding to the target information association data as an abnormal state when the target information association data is greater than a preset threshold; and to determine the target node of the association path corresponding to the target information association data as a normal state when the target information association data is less than or equal to the preset threshold.

[0183] Specific limitations regarding the data mining apparatus can be found in the limitations of the data mining methods described above, and will not be repeated here. Each module in the aforementioned data mining apparatus can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0184] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a data mining method. The display screen can be an LCD screen or an e-ink display. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0185] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0186] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0187] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0188] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, causing the computer device to perform the steps in the above method embodiments.

[0189] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0190] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0191] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0192] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A data mining method, characterized in that, The method includes: Acquire object resource data; wherein, the object resource data is used to characterize object data with multiple dimensions, the object resource data includes user resource data, and the user resource data includes at least user contact information, bank card and enterprise data in multiple dimensions; An object feature graph is constructed based on the object resource data. The object feature graph is used to characterize the node attributes of each node in the object resource data and the association paths between each node. The nodes in the object feature graph include customer nodes, contact information nodes, bank card nodes and enterprise nodes. The edges in the object feature graph include employment relationships and / or bank card binding relationships. Obtain each abnormal node in the object feature map; Calculate the influence factor corresponding to each abnormal node and each other node in the object feature map; wherein, the influence factor is used to characterize the error between the estimated importance value and the true importance value of the abnormal node and other nodes in the object feature map; Each target node among the other nodes is determined based on the influence factors corresponding to each other node, and a corresponding association factor is generated based on the influence factors of each target node. The association factor is used to characterize the degree of association between the target node and the corresponding abnormal node under the corresponding association path. Based on the association paths between the various abnormal nodes and the association factors, target information association data is generated. The target information association data is used to characterize the association relationship between each target node and each abnormal node.

2. The method according to claim 1, characterized in that, The step of calculating the influence factors corresponding to each abnormal node and each other node in the object feature map includes: Obtain the associated paths between each abnormal node and each of the other nodes; Calculate the impact factors of the abnormal node and each other node on each associated path.

3. The method according to claim 1, characterized in that, The step of determining each target node among the other nodes based on the influence factors corresponding to each other node, and generating corresponding correlation factors based on the influence factors of each target node, includes: When the impact factor is greater than a preset threshold, other nodes corresponding to the impact factor are identified as target nodes. Obtain the previous time-series association factor and propagation factor of the target node. The propagation factor is used to characterize the transition probability of the corresponding influence factor between adjacent nodes. The corresponding correlation factor is obtained by fusing the correlation factor, propagation factor, and corresponding influence factor from the previous time step.

4. The method according to claim 1, characterized in that, After determining each target node among the other nodes based on the influence factors corresponding to each other node, and generating corresponding correlation factors based on the influence factors of each target node, the method further includes: Determine the neighboring nodes of the target node, and obtain the adjacent association paths between the target node and the neighboring nodes; The fusion association path between the abnormal node and the adjacent node is obtained based on the adjacent association path and the association path between the abnormal node and the target node. Obtain the previous time-to-date influence factors of the abnormal node and its adjacent nodes; The influence factors of the abnormal node and the target node are fused with the influence factors of the previous moment to obtain the current moment influence factors of the abnormal node and the adjacent nodes. Return to the steps of calculating the influence factors corresponding to each abnormal node and each other node in the object feature map.

5. The method according to claim 4, characterized in that, The step of obtaining the fused association path between the abnormal node and the adjacent node based on the adjacent association path and the association path between the abnormal node and the target node includes: The target node on the path associated with the abnormal node and the target node is taken as the termination node. The merging associated path is generated by concatenating the associated paths using the termination node as the starting point of the adjacent associated paths.

6. The method according to claim 1, characterized in that, The process of generating target information association data by fusing the association paths between the various abnormal nodes and the association factors includes: Obtain the set of association factors corresponding to all associated paths between each abnormal node and each target node; The association factors corresponding to each association path in the association factor set are accumulated and calculated to obtain the target information association data corresponding to the association path.

7. The method according to claim 1, characterized in that, After generating target information association data by fusing the association paths between the various abnormal nodes and the association factors, the method further includes: When the target information associated data is greater than a preset threshold, the target node of the associated path corresponding to the target information associated data is determined to be in an abnormal state. When the target information associated data is less than or equal to a preset threshold, the target node of the associated path corresponding to the target information associated data is determined to be in a normal state.

8. A data mining device, characterized in that, The device includes: A data processing module is used to acquire object resource data; wherein, the object resource data is used to characterize object data with multiple dimensions, the object resource data includes user resource data, the user resource data includes at least user contact information, bank card and enterprise data; an object feature graph is constructed based on the object resource data, the object feature graph is used to characterize the node attributes of each node in the object resource data and the association paths between each node, the nodes in the object feature graph include customer nodes, contact information nodes, bank card nodes and enterprise nodes, and the edges in the object feature graph include employment relationships and / or bank card binding relationships; An impact factor generation module is used to obtain each anomalous node in the object feature map; calculate the impact factor corresponding to each anomalous node and each other node in the object feature map respectively; wherein, the impact factor is used to characterize the error between the estimated importance value and the true importance value of the anomalous node and other nodes in the object feature map; The association factor generation module is used to determine each target node among the other nodes based on the influence factors corresponding to each other node, and to generate corresponding association factors based on the influence factors of each target node. The association factors are used to characterize the degree of association between the target node and the corresponding abnormal node under the corresponding association path. The target information association data generation module is used to generate target information association data by fusing the association paths between the various abnormal nodes and the association factors. The target information association data is used to characterize the association relationship between each target node and each abnormal node.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images