Customer portrait differentiation feature detection method and device, equipment, storage medium and program product

Abstract

Embodiments of the present application provide a customer portrait differentiation feature detection method, device, equipment, storage medium and program product. It belongs to the technical field of big data. It includes: obtaining at least one to-be-processed label corresponding to a target customer identifier; determining a to-be-input matrix according to the to-be-processed label corresponding to each customer identifier; determining a first anomaly score corresponding to a vector in the to-be-input matrix and a first influence value corresponding to a label value by using an isolation forest algorithm; determining a second anomaly score corresponding to the vector in the to-be-input matrix and a second influence value corresponding to the label value by using a local outlier factor algorithm; and determining a differentiation feature of a customer portrait corresponding to the target customer identifier according to the first anomaly score, the second anomaly score, the first influence value corresponding to the target label value of the target vector, and the second influence value corresponding to the target label value. The method solves the problem that the features in the customer portrait are difficult to reflect the differentiation features of customers.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data, and particularly relates to a customer portrait differentiation feature detection method and device, equipment, a storage medium and a program product. BACKGROUND

[0002] At present, in various businesses, in order to better serve customers, the features of customers need to be marked, analyzed or processed.

[0003] At present, in order to more accurately reflect the features of users, the features of customers are usually reflected by establishing a customer portrait.

[0004] However, the customer portrait in the related art can only reflect the regular features of users, and it is difficult to highlight the obvious differences between users. SUMMARY

[0005] The customer portrait differentiation feature detection method and device, equipment, a storage medium and a program product provided by the embodiments of the present application can solve the problem that the customer portrait can only reflect the regular features of users and it is difficult to highlight the obvious differences between users.

[0006] In a first aspect, the embodiments of the present application provide a customer portrait differentiation feature detection method, comprising: obtaining at least one to-be-processed label corresponding to a target customer identifier; determining a to-be-input matrix according to the to-be-processed labels corresponding to each customer identifier, wherein the to-be-input matrix is composed of vectors corresponding to the customer identifiers, and each vector is composed of label values; determining a first anomaly score corresponding to each vector in the to-be-input matrix and a first influence value corresponding to each label value by using an isolation forest algorithm; determining a second anomaly score corresponding to each vector in the to-be-input matrix and a second influence value corresponding to each label value by using a local outlier factor algorithm; and determining a differentiation feature of a customer portrait corresponding to the target customer identifier according to the first anomaly score corresponding to a target vector, the second anomaly score corresponding to the target vector, the first influence value corresponding to a target label value under the target vector, and the second influence value corresponding to the target label value, wherein the target customer identifier corresponds to the target vector.

[0007] In a possible implementation, the determining of the differentiated feature of the customer portrait corresponding to the target customer identifier according to the first anomaly score corresponding to the target vector, the second anomaly score corresponding to the target vector, the first influence value corresponding to the target label value under the target vector, and the second influence value corresponding to the target label value includes: determining the target label value as an abnormal label value if the first influence value or the second influence value corresponding to the target label value is greater than an influence value threshold; determining the occurrence probability of the abnormal label value or a combination of the abnormal label values in the to-be-input matrix; determining the sparsity coefficient corresponding to the abnormal label value or the combination of the abnormal label values according to the occurrence probability; taking the first anomaly score of the target vector as the weight of the first influence value corresponding to the abnormal label value, taking the second anomaly score of the target vector as the weight of the second influence value corresponding to the abnormal label value, and taking the weighted sum of the first influence value and the second influence value as the comprehensive influence value of the abnormal label value; multiplying the comprehensive influence value of the abnormal label value by the sparsity coefficient of the abnormal label value to obtain the differentiated score of the abnormal label value; and determining the differentiated feature of the customer portrait corresponding to the target customer identifier according to the differentiated score of the abnormal label value.

[0008] In a possible implementation, the determining of the differentiated feature of the customer portrait corresponding to the target customer identifier according to the differentiated score of the abnormal label value includes: determining the to-be-processed label corresponding to the abnormal label value as the differentiated feature of the customer portrait corresponding to the target customer identifier if the differentiated score of the abnormal label value of the target customer identifier is greater than or equal to a preset differentiated score threshold; or determining the number of label values of the abnormal label values whose differentiated scores are greater than or equal to the differentiated score threshold under the target customer identifier; if the number of label values is greater than or equal to a preset number threshold, determining the to-be-processed labels corresponding to the M abnormal label values with the largest differentiated scores as the differentiated feature of the customer portrait corresponding to the target customer identifier, where M is a positive integer; or if the differentiated scores of all the abnormal label values of the target customer identifier are all less than the differentiated score threshold, determining the to-be-processed labels corresponding to the N abnormal label values with the largest differentiated scores of the target customer identifier as the differentiated feature of the customer portrait corresponding to the target customer identifier, where N is a positive integer.

[0009] In a possible implementation, after the determining of the differentiated feature of the customer portrait corresponding to the target customer identifier according to the first anomaly score corresponding to the target vector, the second anomaly score corresponding to the target vector, the first influence value corresponding to the target label value under the target vector, and the second influence value corresponding to the target label value, the method further includes: obtaining a business processing result corresponding to the target customer identifier; and changing the differentiated score threshold according to the business processing result.

[0010] In a possible implementation, the sparse coefficient corresponding to the abnormal label value or the combination of abnormal label values is determined according to the occurrence probability, including: subtracting the occurrence probability from a preset value to obtain the sparse coefficient corresponding to the abnormal label value or the combination of abnormal label values; or, according to the occurrence probability, looking up a preset correspondence between the occurrence probability interval and the sparse coefficient to obtain the sparse coefficient corresponding to the abnormal label value or the combination of abnormal label values.

[0011] In a possible implementation, the first abnormal score corresponding to the vector in the input matrix and the first influence value corresponding to the label value are determined by using the Isolation Forest algorithm, including: inputting the input matrix into the Isolation Forest algorithm to obtain a plurality of binary trees divided according to the label value, wherein the binary tree is composed of vectors as nodes; if the target vector is divided into a leaf node of the binary tree by using a target label type corresponding to a target label value, the number of divisions of the target label type corresponding to the target vector is increased by K, where K is a positive integer; the total number of divisions of the target label type corresponding to the target vector is obtained by adding the number of divisions of the target label type corresponding to the target vector in each binary tree; the first influence value corresponding to the target label type is determined according to the total number of divisions of the target label type; the average of the number of edges from the target vector to the root node of each binary tree is determined as the target edge number; and the first abnormal score is determined according to the target edge number.

[0012] In a possible implementation, the second abnormal score corresponding to the vector in the input matrix and the second influence value corresponding to the label value are determined by using the Local Outlier Factor algorithm, including: inputting the input matrix into the Local Outlier Factor algorithm to obtain the second abnormal score corresponding to each vector in the input matrix output by the Local Outlier Factor algorithm; the vector whose second abnormal score is greater than an abnormal score threshold is determined as an outlier vector; the target label value in the outlier vector is replaced with a preset label value, and the step of inputting the input matrix into the Local Outlier Factor algorithm is executed again to obtain a new abnormal factor corresponding to the outlier vector, wherein the label type of the preset label value corresponds to the target label value; and the second influence value of the target label value is determined according to the second abnormal score corresponding to the outlier vector and the new abnormal factor.

[0013] In a second aspect, the embodiments of the present application provide a device for detecting differentiated features of a customer portrait, including:

[0014] The label obtaining module is configured to obtain at least one to-be-processed label corresponding to a target customer identifier; the matrix determining module is configured to determine a to-be-input matrix according to the to-be-processed labels corresponding to each customer identifier, wherein the to-be-input matrix is composed of vectors corresponding to the customer identifiers, and each vector is composed of label values; the first determining module is configured to determine a first anomaly score corresponding to each vector in the to-be-input matrix and a first influence value corresponding to each label value by using an isolation forest algorithm; the second determining module is configured to determine a second anomaly score corresponding to each vector in the to-be-input matrix and a second influence value corresponding to each label value by using a local outlier factor algorithm; and the feature determining module is configured to determine a differentiated feature of a customer portrait corresponding to the target customer identifier according to the first anomaly score corresponding to a target vector, the second anomaly score corresponding to the target vector, the first influence value corresponding to a target label value under the target vector, and the second influence value corresponding to the target label value, wherein the target customer identifier corresponds to the target vector.

[0015] In a third aspect, an embodiment of the present application provides an electronic device, including a memory and a processor; the memory stores computer execution instructions; and the processor executes the computer execution instructions stored in the memory, so that the processor executes the first aspect and / or various possible implementation manners of the first aspect.

[0016] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, which stores computer execution instructions, and the computer execution instructions are executed by a processor to implement the first aspect and / or various possible implementation manners of the first aspect.

[0017] In a fifth aspect, an embodiment of the present application provides a computer program product, which includes a computer program, and the computer program is executed by a processor to implement the first aspect and / or various possible implementation manners of the first aspect.

[0018] The customer portrait differentiated feature detection method, device, equipment, storage medium and program product provided by the embodiment of the present application highlight the characteristics corresponding to the customer identifier by using the isolation forest algorithm and the local outlier factor algorithm to obtain the first anomaly score and the second anomaly score of the vector corresponding to the customer, and the first influence value and the second influence value of the label corresponding to the customer, and obtaining the differentiated feature of the customer portrait corresponding to the target customer identifier according to the first influence value and the second influence value. BRIEF DESCRIPTION OF DRAWINGS

[0019] The accompanying drawings, which are incorporated into and form a part of the specification, illustrate an embodiment consistent with the present application and, together with the specification, serve to explain the principles of the application.

[0020] Figure 1 The scene schematic diagram of the customer portrait differentiated feature detection method provided by the embodiment of the present application is shown in the following figure.

[0021] Figure 2 A flowchart of a method for detecting differentiated features of a customer portrait provided by an embodiment of the present application is shown in FIG. 1.

[0022] Figure 3 A schematic diagram of a use process of an isolation forest algorithm provided by an embodiment of the present application is shown in FIG. 2.

[0023] Figure 4 A structural schematic diagram of a device for detecting differentiated features of a customer portrait provided by an embodiment of the present application is shown in FIG. 3.

[0024] Figure 5 A structural schematic diagram of an electronic device provided by an embodiment of the present application is shown in FIG. 4.

[0025] The specific embodiments of the present application have been shown in the above-described drawings, and will be described in more detail hereinafter. These drawings and the written description are not intended to restrict the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0026] The exemplary embodiments will be described in detail herein with reference to the attached drawings. In the following description, the same numbers are used to indicate the same or similar components. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present application as detailed in the appended claims.

[0027] In the current era of massive business digitization, customer features are the basis for providing quality services.

[0028] Current customer portraits are usually constructed using fixed dimensions, which are difficult to reflect the specific features of customers.

[0029] To solve the above technical problems, the inventors propose the following technical concept: the labels of users are used to form a vector corresponding to a customer, and the vectors corresponding to different customers form a to-be-input matrix. An isolation forest algorithm is used to determine a first anomaly score corresponding to a vector in the to-be-input matrix and a first influence value corresponding to a label. A local outlier factor algorithm is used to determine a second anomaly score corresponding to a vector in the to-be-input matrix and a second influence value corresponding to a label value. The first anomaly score, the first influence value, the second anomaly score, and the second influence value of a target vector are combined to determine a differentiated feature corresponding to a target customer identifier.

[0030] The present application is applied to the scenario of differential feature detection of customer portrait. It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards, and provide corresponding operation portal for user to choose authorization or refusal.

[0031] Figure 1 The scenario schematic diagram of the differential feature detection method of customer portrait provided by the present application. As shown in Figure 1 In this scenario, it includes a server 101 and a data storage unit 102.

[0032] In the specific implementation process, the server 101 can be implemented by using a server or a cluster composed of multiple servers with stronger processing capability and higher security, and in possible cases, a computer, a notebook computer or the like with stronger computing capability can also be used for replacement.

[0033] The data storage unit 102 can include a single database or a combination of multiple databases. For example, one or more of the following databases: Oracle database, MySQL (relational database management system), DRDS (Distribute Relational Database Service, distributed relational database) database, ES (Elasticsearch, elastic search) database, etc., which are not specially limited by the present application.

[0034] The connection between the server 101 and the data storage unit 102 can be wired connection or wireless connection.

[0035] The server 101 is configured to read the customer identification corresponding to the to-be-processed label from the data storage unit 102, generate a to-be-input matrix according to the to-be-processed label, determine a first anomaly score corresponding to a vector in the to-be-input matrix and a first influence value corresponding to a label by using an isolation forest algorithm, determine a second anomaly score corresponding to the vector in the to-be-input matrix and a second influence value corresponding to the label value by using a local outlier factor algorithm, and determine a differential feature corresponding to a target customer identification by combining the first anomaly score, the first influence value, the second anomaly score and the second influence value of the target vector.

[0036] It can be understood that the scenarios shown in the embodiments of the present application do not constitute a specific limitation on the customer portrait differentiated feature detection method. In other feasible embodiments of the present application, the above-mentioned scenarios can include more or fewer components than the illustrations, or combine certain components, or split certain components, or different component arrangements, which can be determined according to actual application scenarios, and are not limited herein. Figure 1 The scenarios shown can be implemented by hardware, software, or a combination of software and hardware.

[0037] The technical solutions of the present application and how the technical solutions of the present application solve the above-mentioned technical problems will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes can not be described again in some embodiments. The embodiments of the present application will be described below with reference to the accompanying drawings.

[0038] Figure 2 The flowchart of the customer portrait differentiated feature detection method provided by the embodiments of the present application is shown. The execution subject of the embodiments of the present application can be the server 101 in Figure 1 , or a computer and / or a mobile phone, etc., and the embodiments of the present application do not make special limitations. As shown in Figure 2 , the method comprises:

[0039] S201: Obtain at least one to-be-processed label corresponding to a target customer identifier.

[0040] In this step, the target customer identifier can be an anonymized identifier of a customer, and the to-be-processed label can be, for example, an address, an age, a type of occupation, etc.

[0041] S202: Determine a to-be-input matrix according to the to-be-processed labels corresponding to each customer identifier, wherein the to-be-input matrix is composed of vectors corresponding to the customer identifiers, and each vector is composed of label values.

[0042] In this step, the to-be-processed label is converted into a numerical value to obtain a label value. The target vector corresponding to the target customer identifier is obtained by combining the label values corresponding to the target customer identifier. The to-be-input matrix is obtained by combining the vectors corresponding to each customer identifier.

[0043] In this step, the to-be-processed label can be converted into a numerical value by using One-Hot encoding, Embedding embedding, etc., or the numerical value of the to-be-processed label can be obtained by using a preset correspondence between the label and the numerical value. The correspondence between the label and the numerical value can be preset by the staff according to experimental data or experience parameters.

[0044] S203: Determine a first anomaly score corresponding to a vector in the to-be-input matrix and a first influence value corresponding to a label value by using an Isolation Forest algorithm.

[0045] This step involves randomly selecting any label type and a random label value to divide the vectors in the input matrix into a left subtree and a right subtree. Then, randomly selecting any label type and a random label value to divide the vectors in the left and right subtrees again, until no further division is possible. This process yields the number of edges from each vector to the root node. Based on the variables from each vector to the root node, the first anomaly score of the vector is determined. Finally, based on the number of times the vector is divided into leaf nodes according to the label value, the first influence value corresponding to the label value is determined.

[0046] Figure 3 This is a schematic diagram illustrating the usage process of the isolated forest algorithm provided in an embodiment of this application. Figure 3 As shown, there are currently 5 vectors forming the input matrix. Based on the label value A, 3 of these 5 vectors are divided into the right subtree, and 2 are divided into the left subtree. The left subtree is further divided using label B, resulting in two leaf nodes. The 3 vectors in the right subtree are further divided using label B, resulting in 2 being divided into the left subtree and 1 into the right subtree (leaf node). Label C is used to divide 1 into the left subtree (leaf node) and 1 into the right subtree (leaf node).

[0047] S204: The local outlier factor algorithm is used to determine the second outlier score corresponding to the vector in the input matrix and the second influence value corresponding to the label value.

[0048] In this step, the input matrix is ​​fed into the Local Outlier Factor Algorithm. The Local Outlier Factor Algorithm treats the vectors in the input matrix as points and calculates the outlier score of each vector, i.e., the second outlier score. By changing the label value in the vector to a preset value or setting it to zero, the outlier score is calculated again. The two outlier scores are subtracted to obtain the second influence value.

[0049] S205: Based on the first abnormal score corresponding to the target vector, the second abnormal score corresponding to the target vector, the first influence value corresponding to the target label value under the target vector, and the second influence value corresponding to the target label value, determine the differentiated features of the customer profile corresponding to the target customer identifier, wherein the target customer identifier corresponds to the target vector.

[0050] This step includes multiplying the first anomaly score by the first impact value to obtain a first product, multiplying the second anomaly score by the second impact value to obtain a second product, and adding the first product and the second product to obtain the total anomaly value. If the total anomaly value is greater than a preset anomaly value threshold, the label to be processed corresponding to the target label value is determined as a differential feature.

[0051] As can be known from the description of the above embodiments, by adopting the Isolation Forest algorithm and the Local Outlier Factor algorithm, the first anomaly score and the second anomaly score of the vector corresponding to the customer are obtained, and the first influence value and the second influence value corresponding to the label of the customer are obtained, and according to the first influence value and the second influence value, the differentiating feature of the customer portrait corresponding to the target customer identifier is obtained, so as to highlight the characteristics corresponding to the customer identifier.

[0052] In a possible implementation, after the step S205, the method further includes: generating the customer portrait corresponding to the target customer identifier according to the preset label and the differentiating feature.

[0053] In a possible implementation, in the step S205, the differentiating feature of the customer portrait corresponding to the target customer identifier is determined according to the first anomaly score corresponding to the target vector, the second anomaly score corresponding to the target vector, the first influence value corresponding to the target label value under the target vector, and the second influence value corresponding to the target label value, and includes:

[0054] S2051: If the first influence value or the second influence value corresponding to the target label value is greater than the influence value threshold, the target label value is determined as an abnormal label value.

[0055] In this step, the influence value threshold can be preset by the staff according to experimental data or experience parameters.

[0056] S2052: Determine the occurrence probability of the abnormal label value or the combination of the abnormal label value in the to-be-input matrix.

[0057] In this step, the number of times of occurrence of the abnormal label value in all vectors is counted, and then divided by the total number of vectors to obtain the occurrence probability. If at least two abnormal label values appear in the target vector, the at least two abnormal label values are taken as a combination of an abnormal label value, the number of times of occurrence of the combination of the abnormal label value in all vectors is counted, and then divided by the total number of vectors to obtain the occurrence probability.

[0058] S2053: Determine the sparsity coefficient corresponding to the abnormal label value or the combination of the abnormal label value according to the occurrence probability.

[0059] In this step, the sparsity coefficient can be obtained by subtracting the occurrence probability from the preset value, or the sparsity coefficient can be obtained by searching the preset correspondence between the occurrence probability range and the sparsity coefficient according to the occurrence probability.

[0060] S2054: Take the first anomaly score of the target vector as the weight of the first influence value corresponding to the abnormal label value, and take the second anomaly score of the target vector as the weight of the second influence value corresponding to the abnormal label value, and sum the weighted first influence value and the weighted second influence value to obtain the comprehensive influence value of the abnormal label value.

[0061] In this step, the first abnormal score is multiplied by the first influence value to obtain a first product, the second abnormal score is multiplied by the second influence value to obtain a second product, and the first product and the second product are added to obtain a comprehensive influence value.

[0062] S2055: multiplying the comprehensive influence value of the abnormal label value by the sparsity coefficient of the abnormal label value to obtain a differentiation score of the abnormal label value.

[0063] In this step, for example, the comprehensive influence value of the abnormal label value is 0.8, the sparsity of the abnormal label value is 0.9, and the differentiation score is 0.72. For another example, the comprehensive influence value of the abnormal label is 0.6, the sparsity is 0.5, and the differentiation score is 0.3.

[0064] S2056: determining a differentiation feature of a customer portrait corresponding to a target customer identifier according to the differentiation score of the abnormal label value.

[0065] In this step, if the differentiation score of the target abnormal label is greater than a preset differentiation score threshold, the to-be-processed label corresponding to the target abnormal label value is determined as the differentiation feature.

[0066] As can be seen from the description of the above embodiments, the embodiments of the present disclosure determine the abnormal label value in the label value according to the influence value, determine the sparsity coefficient of the abnormal label or the abnormal label combination according to the abnormal label value or the probability of the abnormal label value in the to-be-input matrix, take the first abnormal score and the second abnormal score as weights, calculate the weighted sum of the first influence value and the second influence value corresponding to the abnormal label value to obtain a comprehensive influence value, multiply the comprehensive influence value by the sparsity coefficient to obtain a differentiation score, and determine which features are the differentiation features of the customer according to the differentiation score, so as to select more accurate differentiation features by combining the rarity and influence degree of the abnormal label.

[0067] In a possible implementation, in the step S2056, determining a differentiation feature of a customer portrait corresponding to a target customer identifier according to the differentiation score of the abnormal label value, includes: step S2056A, step S2056B or S2056C.

[0068] S2056A: if the differentiation score of the abnormal label value corresponding to the target customer identifier is greater than or equal to a preset differentiation score threshold, the to-be-processed label corresponding to the abnormal label value is determined as the differentiation feature of the customer portrait corresponding to the target customer identifier.

[0069] In this step, for example, under the customer identification A, the differentiated score corresponding to the abnormal label value 1 is 0.8, and the differentiated score threshold is 0.6. Therefore, the to-be-processed label corresponding to the abnormal label value 1 is a differentiated feature. For another example, under the customer identification B, the differentiated score corresponding to the abnormal label value 2 is 0.85, and the differentiated score threshold is 0.7. Therefore, the to-be-processed label corresponding to the abnormal label value 2 is a differentiated feature.

[0070] S2056B: Determine the number of label values of the abnormal label values whose differentiated scores are greater than or equal to the differentiated score threshold under the target customer identification. If the number of label values is greater than or equal to a preset number threshold, determine the to-be-processed labels corresponding to the M abnormal label values with the largest differentiated scores as the differentiated features of the customer portrait corresponding to the target customer identification, where M is a positive integer.

[0071] In this step, for example, under the current customer identification A, the differentiated scores corresponding to 8 abnormal label values are all greater than the differentiated score threshold. Therefore, the to-be-processed labels corresponding to the 4 (M is 4) abnormal label values with the largest differentiated scores are determined as the differentiated features of the customer portrait corresponding to the target customer identification. For another example, under the current customer identification B, the differentiated scores corresponding to 6 abnormal label values are all greater than the differentiated score threshold. Therefore, the to-be-processed labels corresponding to the 3 (M is 3) abnormal label values with the largest differentiated scores are determined as the differentiated features of the customer portrait corresponding to the target customer identification.

[0072] S2056C: If the differentiated scores corresponding to all abnormal label values under the target customer identification are all less than the differentiated score threshold, determine the to-be-processed labels corresponding to the N abnormal label values with the largest differentiated scores under the target customer identification as the differentiated features of the customer portrait corresponding to the target customer identification, where N is a positive integer.

[0073] In this step, N may be 1, 2, or 3, which may be preset by a staff.

[0074] As can be seen from the description of the above embodiments, in the case that there are many abnormal label values with large differentiated scores, all or the M abnormal label values with the largest differentiated scores are selected as the to-be-processed labels as the differentiated features. In the case that there is no abnormal label value greater than the differentiated score threshold, the N abnormal label values with the largest differentiated scores are used as the to-be-processed labels. Therefore, a proper number of to-be-processed labels are selected as the differentiated features.

[0075] After the step S205 determines the differentiated features of the customer portrait corresponding to the target customer identification according to the first abnormal score corresponding to the target vector, the second abnormal score corresponding to the target vector, the first influence value corresponding to the target label value under the target vector, and the second influence value corresponding to the target label value, the method further includes steps S220 and S221.

[0076] S220: Obtain a service processing result corresponding to the target customer identifier.

[0077] In this step, the service processing result can include service processing success or service processing failure.

[0078] S221: Change the differentiated score threshold according to the service processing result.

[0079] In this step, if the service processing fails, the differentiated score threshold is increased, and if the service processing succeeds, the differentiated score threshold is decreased.

[0080] From the description of the above embodiments, it can be known that the embodiments of the present disclosure increase the accuracy of subsequent differentiated labels by dynamically changing the differentiated score threshold according to the service processing result.

[0081] In a possible implementation, the step S2053 of determining the sparsity coefficient corresponding to the abnormal label value or the combination of abnormal label values according to the occurrence probability includes step S2053A or step S2053B.

[0082] S2053A: A preset value is subtracted from the occurrence probability to obtain the sparsity coefficient corresponding to the abnormal label value or the combination of abnormal label values.

[0083] In this step, the preset value is, for example, 1, 0.98, 0.9, etc.

[0084] For example, the preset value is 1, the occurrence probability is 0.2, and the coefficient corresponding to the abnormal label or the combination of abnormal label values is 0.8. For another example, the preset value is 0.9, the occurrence probability is 0.3, and the coefficient corresponding to the abnormal label or the combination of abnormal label values is 0.6.

[0085] S2053B: According to the occurrence probability, the corresponding relationship between the preset occurrence probability interval and the sparsity coefficient is searched to obtain the sparsity coefficient corresponding to the abnormal label value or the combination of abnormal label values.

[0086] In this step, the corresponding relationship between the occurrence probability interval and the sparsity coefficient can be preset by the staff according to experimental data or experience parameters.

[0087] For example, the occurrence probability is 0.73, the corresponding occurrence probability interval is “[0.7, 0.8]”, the occurrence probability interval “[0.7, 0.8]” is used to search the corresponding relationship between the occurrence probability interval and the sparsity coefficient, and the corresponding sparsity coefficient is 0.2. For another example, the occurrence probability is 0.553, the corresponding occurrence probability interval is “[0.5, 0.6]”, the occurrence probability interval “[0.5, 0.6]” is used to search the corresponding relationship between the occurrence probability interval and the sparsity coefficient, and the corresponding sparsity coefficient is 0.4.

[0088] From the description of the above embodiments, the embodiments of the present disclosure make the subsequent differentiated features more accurate by determining the appropriate sparsity coefficient from the occurrence probability.

[0089] In a possible implementation, in the step S203, the first anomaly score corresponding to the vector in the to-be-input matrix and the first influence value corresponding to the label value are determined by using the isolation forest algorithm, including:

[0090] S2031: input the to-be-input matrix into the isolation forest algorithm to obtain a plurality of binary trees divided according to the label value, wherein the binary tree is composed of vectors as nodes.

[0091] In this step, the isolation forest algorithm randomly takes a to-be-processed label type and randomly generates a label value, divides the vectors in the to-be-input matrix into different nodes by using the label value, and obtains a binary tree. By dividing the to-be-input matrix multiple times, a plurality of binary trees are obtained.

[0092] S2032: if the target vector is divided into a leaf node of the binary tree by using a target label type corresponding to a target label value, the division times corresponding to the target label type under the target vector are increased by K, wherein K is a positive integer.

[0093] In this step, for example, when the vectors are divided by using a label type A corresponding to a label value A, vector 1 is divided into a leaf node, and the division times corresponding to the label type A under the vector 1 are increased by 1 (K is 1). For another example, when the vectors are divided by using a label type B corresponding to a label value B, vector 2 is divided into a leaf node, and the division times corresponding to the label type B under the vector 2 are increased by 2 (K is 2).

[0094] S2033: add the division times of the target label type under the target vector in each binary tree to obtain the total division times of the target label type under the target vector.

[0095] In this step, for example, the division times of the target label type of vector A in each binary tree are 2, 1, 3, 1, and 1, and the total division times of the target label type under the target vector are 8. For another example, the division times of the target label type of vector B in each binary tree are 1, 1, 2, 3, and 2, and the total division times of the target label type under the target vector are 9.

[0096] S2034: determine the first influence value corresponding to the target label type according to the total division times of the target label type.

[0097] In this step, the total number of divisions can be directly taken as the first influence value corresponding to the target label type. The total number of divisions can also be divided by the number of binary trees to obtain the first influence value corresponding to the target label type. The total number of divisions can also be mapped to a preset interval to obtain the first influence value.

[0098] For example, vector A is divided into 3 leaf nodes of the binary tree by the target label value, and 3 is determined as the first influence value. For another example, vector B is divided into 5 leaf nodes of the binary tree by the target label value, and 5 is mapped to a preset interval to obtain the first influence value.

[0099] The division threshold can be preset, or can be obtained by calculating the average or mode according to the total number of divisions corresponding to each label type.

[0100] S2035: The average of the number of edges from each target vector to the root node of the binary tree is determined as the target edge number.

[0101] For example Figure 3 In the example, the number of edges from vector "(A4, B4, C4, D4, E4)" to the root node is 3, and the number of edges from vector "(A1, B1, C1, D1, E1)" to the root node is 2.

[0102] S2036: The first anomaly score is determined according to the target edge number.

[0103] In this step, the target edge number can be mapped to the interval of 0 to 1 to obtain the first anomaly score.

[0104] For example, the maximum edge number is mapped to 1, and the minimum edge number is mapped to 0, so that the target edge number is mapped to the first anomaly score.

[0105] As can be seen from the description of the above embodiments, the disclosed embodiments adopt the Isolation Forest algorithm to divide the input matrix into multiple binary trees, determine the influence value corresponding to the target label type according to the number of times the target vector is divided into leaf nodes by the target label value, determine the target edge number by taking the average of the number of edges from the target vector to the root node of the binary tree, and determine the first anomaly score from the target edge number. The disclosed embodiments realize the determination of the abnormal degree of the vector and the influence degree of the label value by the Isolation Forest algorithm, and facilitate the subsequent determination of the differentiated features of the user.

[0106] In one possible implementation, in the step S204, the local outlier factor algorithm is used to determine the second anomaly score of the vector in the input matrix and the second influence value of the label value, including steps S2041 to S2044.

[0107] S2041: inputting the to-be-input matrix into the local outlier factor algorithm to obtain a second abnormal score corresponding to each vector in the to-be-input matrix output by the local outlier factor algorithm.

[0108] In this step, the local outlier factor algorithm determines the second abnormal score corresponding to each vector in the to-be-input matrix by taking each vector in the to-be-input matrix as a point.

[0109] S2042: determining a vector with a second abnormal score greater than an abnormal score threshold value as an outlier vector.

[0110] In this step, the abnormal score threshold value can be set by a worker according to experimental data or an experience parameter in advance.

[0111] S2043: replacing a target label value in the outlier vector with a preset label value, and performing again the step of inputting the to-be-input matrix into the local outlier factor algorithm to obtain a new abnormal factor corresponding to the outlier vector, wherein the preset label value corresponds to a label type of the target label value.

[0112] In this step, the preset label value can be a regular numerical value corresponding to a label value type of the target label value, or can be 0 or a null value, etc. The step of inputting the to-be-input matrix into the local outlier factor algorithm can be the step S2041 described above.

[0113] S2044: determining a second influence value of the target label value according to the second abnormal score corresponding to the outlier vector and the new abnormal factor.

[0114] In this step, the second influence value of the target label value can be obtained by calculating a difference between the second abnormal score corresponding to the outlier vector and the new abnormal factor. Alternatively, the second influence value of the target label value can be obtained by calculating a difference between the second abnormal score and the new abnormal factor, multiplying the difference by a preset value, and then obtaining the second influence value of the target label value.

[0115] The second influence value of each target label value can be obtained by repeating the steps described above.

[0116] As can be seen from the description of the above embodiments, the second abnormal score corresponding to each vector is obtained by using the local outlier factor algorithm, the numerical value of the target label value is changed, the second abnormal score of the vector is calculated again, the second influence value corresponding to the target label value is determined according to the change degree of the second abnormal score calculated twice, and thus the difference feature of the user is determined subsequently.

[0117] Figure 4 The structure schematic diagram of the customer portrait difference feature detection device provided by the embodiments of the present application is shown in FIG. 1. Figure 4As shown, the customer portrait differentiation feature detection apparatus 400 comprises a label acquisition module 401, a matrix determination module 402, a first determination module 403, a second determination module 404, and a feature determination module 405.

[0118] The label acquisition module 401 is configured to acquire at least one to-be-processed label corresponding to a target customer identifier.

[0119] The matrix determination module 402 is configured to determine a to-be-input matrix according to the to-be-processed labels corresponding to each customer identifier, wherein the to-be-input matrix is composed of vectors corresponding to the customer identifiers, and each vector is composed of label values.

[0120] The first determination module 403 is configured to determine, by using an isolation forest algorithm, a first anomaly score corresponding to each vector in the to-be-input matrix and a first influence value corresponding to each label value.

[0121] The second determination module 404 is configured to determine, by using a local outlier factor algorithm, a second anomaly score corresponding to each vector in the to-be-input matrix and a second influence value corresponding to each label value.

[0122] The feature determination module 405 is configured to determine, according to the first anomaly score corresponding to a target vector, the second anomaly score corresponding to the target vector, the first influence value corresponding to a target label value under the target vector, and the second influence value corresponding to the target label value, a differentiation feature of a customer portrait corresponding to a target customer identifier, wherein the target customer identifier corresponds to the target vector.

[0123] The apparatus provided in the embodiment can be used to execute the technical solutions of the method embodiments, and has similar principles and technical effects, which will not be described here again.

[0124] In a possible implementation, the feature determination module 405 is configured to determine, if the first influence value or the second influence value corresponding to a target label value is greater than an influence value threshold, the target label value as an abnormal label value; determine an occurrence probability of the abnormal label value or a combination of abnormal label values in the to-be-input matrix; determine a sparsity coefficient corresponding to the abnormal label value or the combination of abnormal label values according to the occurrence probability; take the first anomaly score of the target vector as a weight of the first influence value corresponding to the abnormal label value, take the second anomaly score of the target vector as a weight of the second influence value corresponding to the abnormal label value, and obtain a comprehensive influence value of the abnormal label value by weighting and summing the first influence value and the second influence value; multiply the comprehensive influence value of the abnormal label value by the sparsity coefficient of the abnormal label value to obtain a differentiation score of the abnormal label value; and determine the differentiation feature of the customer portrait corresponding to the target customer identifier according to the differentiation score of the abnormal label value.

[0125] In a possible implementation, the feature determination module 405 is configured to: if the difference score corresponding to the abnormal label value of the target customer identification is greater than or equal to a preset difference score threshold, determine that the to-be-processed label corresponding to the abnormal label value is a difference feature of the customer portrait corresponding to the target customer identification; or, determine the number of abnormal label values whose difference scores are greater than or equal to the difference score threshold under the target customer identification; if the number of abnormal label values is greater than or equal to a preset number threshold, determine the M abnormal label values corresponding to the to-be-processed labels with the largest difference scores as the difference features of the customer portrait corresponding to the target customer identification, where M is a positive integer; or, if the difference scores corresponding to all abnormal label values under the target customer identification are all less than the difference score threshold, determine the N abnormal label values corresponding to the to-be-processed labels with the largest difference scores of the target customer identification as the difference features of the customer portrait corresponding to the target customer identification, where N is a positive integer.

[0126] In a possible implementation, the feature determination module 405 is configured to: subtract the occurrence probability from the preset value to obtain a sparsity coefficient corresponding to the abnormal label value or the combination of abnormal label values; or, according to the occurrence probability, find a correspondence between a preset occurrence probability interval and a sparsity coefficient to obtain a sparsity coefficient corresponding to the abnormal label value or the combination of abnormal label values.

[0127] In a possible implementation, the first determination module 403 is specifically configured to: input the to-be-input matrix into the isolation forest algorithm to obtain a plurality of binary trees divided according to label values, where each binary tree is composed of vectors as nodes; if the target vector is divided into leaf nodes of the binary tree by using the target label type corresponding to the target label value, increase the division times of the target label type corresponding to the target vector by K, where K is a positive integer; add the division times of the target label type corresponding to the target vector in each binary tree to obtain the total division times of the target label type corresponding to the target vector; determine a first influence value corresponding to the target label type according to the total division times of the target label type; determine the target edge number as an average value of the number of edges from the target vector to the root node of each binary tree; and determine the first abnormal score according to the target edge number.

[0128] In a possible implementation, the second determination module 404 is specifically configured to: input the to-be-input matrix into the local outlier factor algorithm to obtain a second abnormal score corresponding to each vector in the to-be-input matrix output by the local outlier factor algorithm; determine a vector whose second abnormal score is greater than an abnormal score threshold as an outlier vector; replace the target label value in the outlier vector with a preset label value, and execute again the step of inputting the to-be-input matrix into the local outlier factor algorithm to obtain a new abnormal factor corresponding to the outlier vector, where the label type of the preset label value corresponds to that of the target label value; and determine a second influence value of the target label value according to the second abnormal score corresponding to the outlier vector and the new abnormal factor.

[0129] The device provided by the embodiment can be used to execute the technical solutions of the method embodiments, and has similar implementation principles and technical effects, which will not be described here again.

[0130] To implement the above embodiments, the embodiment of the present application further provides an electronic device.

[0131] Reference Figure 5 , which shows a structural schematic diagram of an electronic device 500 suitable for implementing the embodiments of the present application. The electronic device 500 can be a terminal device or a server. The terminal device can include, but is not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, personal digital assistants (PDA), tablet computers (Portable Android Device, PAD), portable multimedia players (Portable Media Player, PMP), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), and the like, and fixed terminals such as digital TVs, desktop computers, and the like. Figure 5 The electronic device shown is only an example and should not limit the functions and use range of the embodiments of the present application.

[0132] As Figure 5 shown, the electronic device 500 can include a processor (such as a central processor, a graphics processor, etc.) 501, and a memory 502 connected in communication with the processor, which can perform various appropriate actions and processes according to the program, computer execution instructions stored in the memory 502 or the program loaded from the storage device 508 into the random access memory (Random Access Memory, RAM) 503, to implement the customer portrait differentiation feature detection method in any of the above embodiments. The memory can be a read-only memory (Read Only Memory, ROM). In the RAM 503, various programs and data required for the operation of the electronic device 500 are also stored. The processing device 501, the memory 502, and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0133] Generally, the following devices can be connected to the I / O interface 505: input devices 506 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, and the like; output devices 507 including, for example, a Liquid Crystal Display (LCD), a speaker, a vibrator, and the like; storage devices 508 including, for example, a tape, a hard disk, and the like; and communication devices 509. The communication devices 509 can allow the electronic device 500 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 The electronic device 500 is shown with various devices, but it is understood that all of the shown devices are not required to be implemented or present. More or less devices can alternatively be implemented or present.

[0134] In particular, the processes described above with reference to the flowcharts can be implemented as a computer software program according to embodiments of the present application. For example, embodiments of the present application include a computer program product comprising a computer program carried on a computer-readable storage medium, the computer program containing program code for executing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication devices 509, or installed from the storage devices 508, or installed from the memory 502. When the computer program is executed by the processing devices 501, the above-mentioned functions defined in the methods of embodiments of the present application are performed.

[0135] It should be noted that the computer-readable storage medium in the embodiments of the present application can be a computer-readable signal medium or a computer storage medium, or any combination thereof. The computer-readable storage medium may, for example, be, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or apparatus, or any combination thereof. More specific examples of the computer-readable storage medium include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present application, the computer-readable storage medium can be any tangible medium that contains or stores a program used by or in connection with an instruction execution system, apparatus, or device. In the present application, the computer-readable signal medium can include a data signal propagated in a baseband or as a carrier wave in a propagated data signal, which bears computer-readable program code. Such a propagated data signal can take many forms, including but not limited to, an electromagnetic signal, an optical signal, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable storage medium that can send, propagate, or transfer a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained in the computer-readable storage medium can be transmitted by any suitable medium, including but not limited to, wire, cable, RF (radio frequency), or the like, or any suitable combination thereof.

[0136] The computer-readable storage medium described above can be included in the electronic device described above; or can exist separately from the electronic device and be not assembled in the electronic device.

[0137] The computer-readable storage medium described above carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the methods shown in the embodiments described above.

[0138] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0139] The computer program instructions can also be loaded onto a computer or other programmable information processing apparatus to cause a series of operations to be performed on the computer or other programmable information processing apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable information processing apparatus implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0140] The modules involved in the embodiments described in the present application can be implemented in software or hardware. Among them, the name of the unit does not constitute a limitation to the module itself in some cases, for example, the first determination module can also be described as "the first influence value determination module".

[0141] The functions described above in the specification can be performed at least in part by one or more hardware logic components. For example, and without limitation, illustrative types of hardware logic components that can be used include Field-programmable Gate Arrays (FPGAs), Program-specific Integrated Circuits (ASICs), Program-specific Standard Products (ASSPs), System-on-a-chip systems (SOCs), Complex Programmable Logic Devices (CPLDs), etc.

[0142] The application further provides a computer readable storage medium, which stores computer execution instructions, and when a processor executes the computer execution instructions, the technical solution of the customer portrait differentiated feature detection method in any of the above embodiments is implemented, the implementation principle and beneficial effects of which are similar to those of the customer portrait differentiated feature detection method, and can be referred to the implementation principle and beneficial effects of the customer portrait differentiated feature detection method, which will not be described herein again.

[0143] In the context of the present application, the machine readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus or device. The machine readable medium can be a machine readable signal medium or a machine readable storage medium. The machine readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any suitable combination of the above. More specific examples of the machine readable storage medium can include one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical storage devices, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0144] The application further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the technical solution of the customer portrait differentiated feature detection method in any of the above embodiments is implemented, the implementation principle and beneficial effects of which are similar to those of the customer portrait differentiated feature detection method, and can be referred to the implementation principle and beneficial effects of the customer portrait differentiated feature detection method, which will not be described herein again.

[0145] The above description is merely preferred embodiments of the present application and a description of the principles of the applied technology. Those skilled in the art should understand that the disclosed range of the present application is not limited to the technical solutions formed by the specific combinations of the above technical features, and also covers other technical solutions formed by any combinations of the above technical features or equivalent features without departing from the disclosed concept. For example, the above features can be replaced with other technical features disclosed in the present application (but not limited to) having similar functions to form technical solutions.

[0146] Those skilled in the art can understand that all or part of the steps of the above-mentioned method embodiments can be completed by program instruction related hardware. The foregoing program can be stored in a computer readable storage medium. The program executes to perform the steps of the above-mentioned method embodiments; and the foregoing storage medium includes various media capable of storing program codes, such as ROM, RAM, magnetic disk, or optical disk.

[0147] Finally, it should be noted that other embodiments of the present application will readily occur to those skilled in the art upon consideration of the specification and practice of the present application disclosed herein. The present application is intended to include all such variations as fall within the general scope of the application, and includes the generic principles disclosed and the best mode known to the inventors to be currently practiced as well as variations thereof, without departing from the scope of the present application as defined by the claims. The specification and examples give the best application of the present application as currently known, and together with the description of the application serve to best illustrate the principles of the application. The scope of the application is expressly set forth in the claims.

Claims

1. A method for detecting differentiated features of a customer portrait, characterized in that, The method comprises: obtaining at least one to-be-processed label corresponding to a target customer identifier; determining a to-be-input matrix corresponding to each customer identifier according to the to-be-processed label of the customer identifier, wherein the to-be-input matrix is composed of vectors corresponding to the customer identifier, and each vector is composed of label values; determining a first anomaly score corresponding to the vector and a first influence value corresponding to the label value in the to-be-input matrix by using an isolation forest algorithm, comprising: inputting the to-be-input matrix into the isolation forest algorithm to obtain a plurality of binary trees divided according to the label value, wherein each binary tree is composed of vectors as nodes; if a target vector is divided into a leaf node of the binary tree by using a target label type corresponding to a target label value, then the number of divisions of the target label type under the target vector is increased by K, wherein K is a positive integer; adding the number of divisions of the target label type under the target vector in each binary tree to obtain a total number of divisions of the target label type under the target vector; determining the first influence value corresponding to the target label type according to the total number of divisions of the target label type; determining a target edge number according to an average value of the number of edges from the target vector to the root node of each binary tree; and determining the first anomaly score according to the target edge number; determining a second anomaly score corresponding to the vector and a second influence value corresponding to the label value in the to-be-input matrix by using a local outlier factor algorithm; determining a differentiation feature of a customer portrait corresponding to a target customer identifier according to a first anomaly score corresponding to a target vector, a second anomaly score corresponding to the target vector, a first influence value corresponding to a target label value under the target vector, and a second influence value corresponding to the target label value, wherein the target customer identifier corresponds to the target vector, comprising: if the first influence value or the second influence value corresponding to the target label value is greater than an influence value threshold, then determining the target label value as an abnormal label value; determining an occurrence probability of the abnormal label value or a combination of the abnormal label values in the to-be-input matrix; determining a sparsity coefficient corresponding to the abnormal label value or the combination of the abnormal label values according to the occurrence probability; taking the first anomaly score of the target vector as a weight of the first influence value corresponding to the abnormal label value, taking the second anomaly score of the target vector as a weight of the second influence value corresponding to the abnormal label value, and taking a weighted sum of the first influence value and the second influence value to obtain a comprehensive influence value of the abnormal label value; multiplying the comprehensive influence value of the abnormal label value by the sparsity coefficient of the abnormal label value to obtain a differentiation score of the abnormal label value; and determining the differentiation feature of the customer portrait corresponding to the target customer identifier according to the differentiation score of the abnormal label value.

2. The method of claim 1, wherein, The determination of the differentiation feature of the customer portrait corresponding to the target customer identifier according to the differentiation score of the abnormal label value comprises: If the difference score corresponding to the abnormal label value of the target customer identifier is greater than or equal to the preset difference score threshold, the processed label corresponding to the abnormal label value is determined as a difference feature of the customer portrait corresponding to the target customer identifier; or, The number of label values of abnormal label values with a difference score greater than or equal to the difference score threshold under the target customer identifier is determined; if the number of label values is greater than or equal to a preset number threshold, the M abnormal label values with the largest difference score correspond to the processed label determined as the difference feature of the customer portrait corresponding to the target customer identifier, where M is a positive integer; or, If the difference score corresponding to all abnormal label values under the target customer identifier is less than the difference score threshold, the N abnormal label values corresponding to the processed label with the largest difference score are determined as the difference feature of the customer portrait corresponding to the target customer identifier, where N is a positive integer.

3. The method of claim 1, wherein, The difference score corresponding to the abnormal label value or the combination of abnormal label values is determined according to the occurrence probability, including: The difference score corresponding to the abnormal label value or the combination of abnormal label values is obtained by subtracting the occurrence probability from a preset value; or, According to the occurrence probability, the corresponding relationship between the preset occurrence probability interval and the sparsity coefficient is found, and the difference score corresponding to the abnormal label value or the combination of abnormal label values is obtained.

4. The method according to any one of claims 1 to 3, characterized in that, The second abnormal score corresponding to the vector in the input matrix and the second influence value corresponding to the label value are determined by using the local outlier factor algorithm, including: The input matrix is input into the local outlier factor algorithm to obtain the second abnormal score corresponding to each vector in the input matrix output by the local outlier factor algorithm; The vector with a second abnormal score greater than an abnormal score threshold is determined as an outlier vector; The target label value in the outlier vector is replaced by a preset label value, and the step of inputting the input matrix into the local outlier factor algorithm is executed again to obtain a new abnormal factor corresponding to the outlier vector, wherein the preset label value corresponds to the label type of the target label value; The second influence value of the target label value is determined according to the second abnormal score and the new abnormal factor corresponding to the outlier vector.

5. A device for detecting differentiated features of a customer profile, characterized in that, It includes: The label acquisition module is used for acquiring at least one processed label corresponding to the target customer identifier; The matrix determination module is used for determining an input matrix according to the processed label corresponding to each customer identifier, wherein the input matrix is composed of vectors corresponding to customer identifiers, and the vector is composed of label values; The first determining module is configured to determine the first anomaly score corresponding to the vector and the first influence value corresponding to the label value in the input matrix by using an isolation forest algorithm, including: inputting the input matrix into the isolation forest algorithm to obtain a plurality of binary trees divided according to the label value, wherein the binary tree is composed of the vector as a node; if a target vector is divided into a leaf node of the binary tree by using a target label type corresponding to a target label value, then the number of divisions of the target label type under the target vector is increased by K, wherein K is a positive integer; the number of divisions of the target label type under the target vector in each binary tree is added to obtain the total number of divisions of the target label type under the target vector; the first influence value corresponding to the target label type is determined according to the total number of divisions of the target label type; the average value of the number of edges from the target vector to the root node of each binary tree is determined as a target edge number; and the first anomaly score is determined according to the target edge number; The second determining module is configured to determine the second anomaly score corresponding to the vector and the second influence value corresponding to the label value in the input matrix by using a local outlier factor algorithm. The feature determining module is configured to determine the differential feature of the customer portrait corresponding to the target customer identifier according to the first anomaly score corresponding to the target vector, the second anomaly score corresponding to the target vector, the first influence value corresponding to the target label value under the target vector, and the second influence value corresponding to the target label value, wherein the target customer identifier corresponds to the target vector, including: if the first influence value or the second influence value corresponding to the target label value is greater than an influence value threshold, then the target label value is determined as an abnormal label value; the probability of occurrence of the abnormal label value or a combination of the abnormal label values in the input matrix is determined; the sparsity coefficient corresponding to the abnormal label value or the combination of the abnormal label values is determined according to the probability of occurrence; the first anomaly score of the target vector is used as the weight of the first influence value corresponding to the abnormal label value, the second anomaly score of the target vector is used as the weight of the second influence value corresponding to the abnormal label value, the first influence value and the second influence value are weighted and summed to obtain a comprehensive influence value of the abnormal label value; the differential score of the abnormal label value is obtained by multiplying the comprehensive influence value of the abnormal label value and the sparsity coefficient of the abnormal label value; and the differential feature of the customer portrait corresponding to the target customer identifier is determined according to the differential score of the abnormal label value.

6. An electronic device, comprising: The memory and the processor are included. The memory stores computer execution instructions. The processor executes the computer execution instructions stored in the memory, so that the processor executes the method of any one of claims 1 to 4. The computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by the processor to implement the method of any one of claims 1 to 4.

7. A computer readable storage medium characterized by ​ 8. A computer program product, characterised in that, A computer program comprising computer program elements which, when executed by a processor, perform the method of any one of claims 1 to 4.

Images