Customer feature positioning method, device, equipment and storage medium

Abstract

The present invention relates to artificial intelligence, and provides a method, device, equipment and storage medium for locating customer characteristics. The method can obtain dimension information of preset customers on multiple preset dimensions, generate a customer characterization matrix according to the dimension mapping table and dimension information of each preset dimension, perform predictive analysis on the customer characterization matrix based on a category prediction model, obtain user categories and category probabilities, screen target customers from preset customers according to user categories and category probabilities, extract feature matrices from the customer characterization matrix of target customers based on each preset dimension, generate the influence of each preset dimension on the target customer according to the predicted probability and category probability of each feature matrix on the user category according to the category prediction model, generate feature information of the target user according to the influence and user category, and accurately generate feature information. In addition, the present invention also relates to blockchain technology, and the feature information can be stored in the blockchain.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a customer feature positioning method, device, equipment and storage medium. Background Art

[0002] In order to take targeted business development measures for customers, it is necessary to locate the key features that affect customer labels. However, in the current process of locating key features, customer features are usually analyzed directly based on dimension weights. Since this method does not take into account the impact of the combination of multiple customer features on customer labels, it is impossible to accurately locate key features, which is not conducive to business development. Summary of the invention

[0003] In view of the above, it is necessary to provide a customer feature positioning method, device, equipment and storage medium, which can accurately locate the feature information that affects the user category.

[0004] In one aspect, the present invention provides a method for locating customer characteristics, the method comprising:

[0005] Obtain dimension information of a preset customer on multiple preset dimensions;

[0006] Generate a customer representation matrix for the preset customer according to the dimension mapping table of each preset dimension and the dimension information;

[0007] Performing prediction analysis on the customer representation matrix based on a pre-trained category prediction model to obtain user categories and category probabilities of the user categories;

[0008] Screening target customers from the preset customers according to the user category and the category probability;

[0009] Extracting a feature matrix from the customer representation matrix of the target customer based on each preset dimension;

[0010] Generate the influence of each preset dimension on the target customer according to the predicted probability of each feature matrix on the user category and the category probability of the category prediction model;

[0011] The characteristic information of the target user is generated according to the influence and the user category.

[0012] According to a preferred embodiment of the present invention, generating the customer representation matrix of the preset customer according to the dimension mapping table of each preset dimension and the dimension information includes:

[0013] Acquire a list corresponding to each preset dimension from a preset mapping library to obtain a plurality of dimension mapping tables;

[0014] Based on each dimension mapping table, corresponding dimension information is mapped in parallel to obtain a dimension vector of each preset dimension;

[0015] Obtaining a weight matrix of any configuration category in the category prediction model;

[0016] Identify the weight position of the dimension weight of each preset dimension in the weight matrix;

[0017] The dimensional vectors are concatenated according to the weight positions to obtain the customer representation matrix.

[0018] According to a preferred embodiment of the present invention, the category prediction model includes a normalization layer and an output layer, and the customer representation matrix is ​​predicted and analyzed based on the pre-trained category prediction model to obtain user categories and category probabilities of the user categories, including:

[0019] Calculate the product of each weight matrix and the customer characterization matrix to obtain the output score of the preset customer in each configuration category;

[0020] The output scores are standardized based on the standardization layer to obtain the output probability of the preset customer in each configuration category. The calculation formula of the output probability is:

[0021]

[0022] Among them, P i refers to the output probability of the i-th configuration category, x i refers to the output score of the i-th configuration category, μ refers to the average output score of all configuration categories in the category prediction model, and σ refers to the standard deviation of the output scores of all configuration categories;

[0023] Obtaining a preset probability threshold for each configuration category from the output layer;

[0024] An output probability whose value is greater than a corresponding preset probability threshold is determined as the category probability, and a configuration category corresponding to the category probability is determined as the user category.

[0025] According to a preferred embodiment of the present invention, generating the influence of each preset dimension on the target customer based on the predicted probability of each feature matrix on the user category and the category probability according to the category prediction model includes:

[0026] Input each feature matrix into the category prediction model, and extract the output value of each feature matrix on the user category from the normalization layer as the prediction probability;

[0027] The difference between each predicted probability and the category probability is calculated to obtain the influence degree.

[0028] According to a preferred embodiment of the present invention, extracting a feature matrix from a customer characterization matrix of the target customer based on each preset dimension comprises:

[0029] For any preset dimension, the dimension vector corresponding to any preset dimension in the customer representation matrix of the target customer is deleted to obtain the feature matrix.

[0030] According to a preferred embodiment of the present invention, screening target customers from the preset customers according to the user category and the category probability includes:

[0031] Screening out preset customers corresponding to the user category as customers to be tested;

[0032] Counting the number of customers of the customer to be tested, and calculating the product of the number of customers and a preset ratio to obtain a target number;

[0033] Sort the preset customers according to the category probabilities of the customers to be tested in the user categories to obtain a customer list;

[0034] The target customers are selected from the customer list in sequence according to the target quantity.

[0035] According to a preferred embodiment of the present invention, generating the characteristic information of the target user according to the influence and the user category includes:

[0036] Determine the preset dimension with the greatest influence as the target dimension;

[0037] Extracting information of the target user on the target dimension from the dimension information as a target factor;

[0038] The characteristic information is determined according to the target factor and the user category.

[0039] On the other hand, the present invention further provides a customer feature locating device, the customer feature locating device comprising:

[0040] An acquisition unit, used for acquiring dimension information of a preset customer on multiple preset dimensions;

[0041] A generating unit, configured to generate a customer representation matrix of the preset customer according to a dimension mapping table of each preset dimension and the dimension information;

[0042] A prediction unit, configured to perform prediction analysis on the customer representation matrix based on a pre-trained category prediction model to obtain user categories and category probabilities of the user categories;

[0043] A screening unit, used for screening target customers from the preset customers according to the user category and the category probability;

[0044] An extraction unit, configured to extract a feature matrix from a customer representation matrix of the target customer based on each preset dimension;

[0045] The generating unit is further used to generate the influence of each preset dimension on the target customer according to the predicted probability of each feature matrix on the user category and the category probability of the category prediction model;

[0046] The generating unit is further configured to generate characteristic information of the target user according to the influence and the user category.

[0047] On the other hand, the present invention further provides an electronic device, comprising:

[0048] a memory storing computer readable instructions; and

[0049] A processor executes the computer-readable instructions stored in the memory to implement the customer feature positioning method.

[0050] On the other hand, the present invention further proposes a computer-readable storage medium, in which computer-readable instructions are stored. The computer-readable instructions are executed by a processor in an electronic device to implement the customer feature positioning method.

[0051] It can be seen from the above technical solutions that the present invention constructs the customer characterization matrix through a dimension mapping table of each preset dimension. Since each dimension mapping table only stores mapping information of the corresponding preset dimension, the construction efficiency of the customer characterization matrix can be improved. The customer characterization matrix is ​​predicted and analyzed based on the category prediction model. Since the customer characterization matrix combines the dimension information on the multiple preset dimensions, the prediction accuracy of the user category and the category probability can be improved. Then, the target customers can be screened according to the category probability and the user category, and customers with analytical significance can be accurately selected, thereby improving the success rate of business personnel in business development. Furthermore, based on the multiple preset dimensions, the feature matrix can be quickly extracted from the customer characterization matrix of the target customer, thereby improving the generation efficiency of the feature information. In addition, combined with the predicted probability of the feature matrix on the user category and the category probability of the customer characterization matrix of the target customer on the user category, the influence of each preset dimension on the target customer can be accurately determined, thereby improving the positioning accuracy of the feature information. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1It is a flow chart of a preferred embodiment of the customer feature positioning method of the present invention.

[0053] Figure 2 It is a functional module diagram of a preferred embodiment of the customer feature locating device of the present invention.

[0054] Figure 3 It is a schematic diagram of the structure of an electronic device of a preferred embodiment of the method for locating customer characteristics of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0056] like Figure 1 FIG. 1 is a flowchart of a preferred embodiment of the method for locating customer characteristics of the present invention. According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted.

[0057] The customer feature positioning method can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0058] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0059] The customer feature positioning method is applied to one or more electronic devices, which are devices that can automatically perform numerical calculations and / or information processing according to pre-set or stored computer-readable instructions, and whose hardware includes but is not limited to microprocessors, application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), digital signal processors (DSP), embedded devices, etc.

[0060] The electronic device may be any electronic product that can perform human-computer interaction with a user, such as a personal computer, a tablet computer, a smart phone, a personal digital assistant (PDA), a game console, an interactive network television (IPTV), a smart wearable device, etc.

[0061] The electronic device may include a network device and / or a user device. The network device includes, but is not limited to, a single network electronic device, an electronic device group consisting of multiple network electronic devices, or a cloud consisting of a large number of hosts or network electronic devices based on cloud computing.

[0062] The network where the electronic device is located includes, but is not limited to: the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (VPN), etc.

[0063] S10, obtaining dimension information of a preset customer on multiple preset dimensions.

[0064] In at least one embodiment of the present invention, the preset customer refers to a purchasing user of a certain product.

[0065] The multiple preset dimensions include, but are not limited to: customer basic information category, customer transaction behavior category, customer contact behavior category, etc.

[0066] The dimension information refers to the relevant information of the preset customer on the multiple preset dimensions. The dimension information can be obtained from a customer relationship management system.

[0067] S11, generating a customer representation matrix of the preset customer according to the dimension mapping table of each preset dimension and the dimension information.

[0068] In at least one embodiment of the present invention, the dimension mapping table stores a mapping relationship between data information on a corresponding preset dimension and a vector. The vector dimension in the dimension mapping table can be 100 dimensions or 300 dimensions. The present invention does not limit the vector dimension of the dimension mapping table.

[0069] The customer characterization matrix is ​​used to indicate the dimension information of the preset customer on the multiple preset dimensions.

[0070] In at least one embodiment of the present invention, the electronic device generates a customer representation matrix of the preset customer according to the dimension mapping table of each preset dimension and the dimension information, including:

[0071] Acquire a list corresponding to each preset dimension from a preset mapping library to obtain a plurality of dimension mapping tables;

[0072] Based on each dimension mapping table, corresponding dimension information is mapped in parallel to obtain a dimension vector of each preset dimension;

[0073] Obtaining a weight matrix of any configuration category in the category prediction model;

[0074] Identify the weight position of the dimension weight of each preset dimension in the weight matrix;

[0075] The dimensional vectors are concatenated according to the weight positions to obtain the customer representation matrix.

[0076] The preset mapping library stores mapping lists of multiple dimensions, and each mapping list stores a mapping relationship between data information and vectors in a corresponding dimension.

[0077] The weight matrix is ​​a matrix generated when training the category prediction model, and the weight matrix is ​​used to quantify the importance of the corresponding preset dimension to any of the configuration categories. It is understandable that in the category prediction model, the weight matrix corresponding to each configuration category is different, and the importance of the preset dimensions to the configuration category in each weight matrix is ​​spliced ​​in the same way. For example, the first row vector in the weight matrix is ​​the importance of the first preset dimension to the configuration category, the second row vector in the weight matrix is ​​the importance of the second preset dimension to the configuration category, and so on.

[0078] By mapping the dimensional information in parallel through the multiple dimensional mapping tables, the generation efficiency of the dimensional vector is improved, and then the dimensional vector is spliced ​​based on the weight position of the dimensional weight in the weight matrix, which can ensure the subsequent prediction accuracy of the user category and the category probability, and improve the rationality of the generation of the customer characterization matrix.

[0079] S12, performing prediction analysis on the customer representation matrix based on a pre-trained category prediction model to obtain user categories and category probabilities of the user categories.

[0080] In at least one embodiment of the present invention, the category prediction model is used to predict the specific category of the customer and the probability that the customer is in the specific category.

[0081] The user categories may include multiple ones. For example, the user category of customer A may be easy to add insurance, easy to refer, and difficult to pay.

[0082] The category probability refers to the probability value that the category prediction model predicts that the preset customer is the user category.

[0083] In at least one embodiment of the present invention, the category prediction model includes a normalization layer and an output layer, and the electronic device performs a prediction analysis on the customer representation matrix based on a pre-trained category prediction model to obtain user categories and category probabilities of the user categories, including:

[0084] Calculate the product of each weight matrix and the customer characterization matrix to obtain the output score of the preset customer in each configuration category;

[0085] The output scores are standardized based on the standardization layer to obtain the output probability of the preset customer in each configuration category. The calculation formula of the output probability is:

[0086]

[0087] Among them, P i refers to the output probability of the i-th configuration category, x i refers to the output score of the i-th configuration category, μ refers to the average output score of all configuration categories in the category prediction model, and P refers to the standard deviation of the output scores of all configuration categories;

[0088] Obtaining a preset probability threshold for each configuration category from the output layer;

[0089] An output probability whose value is greater than a corresponding preset probability threshold is determined as the category probability, and a configuration category corresponding to the category probability is determined as the user category.

[0090] The preset probability threshold is generated when training the category prediction model based on multiple training samples.

[0091] By analyzing the customer characterization matrix through the weight matrix, the output score can be accurately determined, and then the user category can be determined based on the preset probability threshold. Since the preset probability threshold is generated when the category prediction model is trained after multiple training samples, the prediction accuracy of the user category can be improved.

[0092] S13, screening target customers from the preset customers according to the user category and the category probability.

[0093] In at least one embodiment of the present invention, the target customer refers to a preset customer with the user category, and the category probability of the preset customer in the user category is ranked in the top N or bottom N, where N refers to the number of the target customers.

[0094] In at least one embodiment of the present invention, the electronic device screening target customers from the preset customers according to the user category and the category probability comprises:

[0095] Screening out preset customers corresponding to the user category as customers to be tested;

[0096] Counting the number of customers of the customer to be tested, and calculating the product of the number of customers and a preset ratio to obtain a target number;

[0097] Sort the preset customers according to the category probabilities of the customers to be tested in the user categories to obtain a customer list;

[0098] The target customers are selected from the customer list in sequence according to the target quantity.

[0099] The customer to be tested refers to a preset customer with the user category.

[0100] The preset ratio can be set according to demand, for example, the preset ratio can be 5%.

[0101] The arrangement order of the customer list may be arranged in the order from large to small according to the category probability, or may be arranged in the order from small to large according to the category probability.

[0102] The order of selecting the target customers corresponds to the order of arranging the customer list. For example, if the order of arranging the customer list is from large to small, then the order of selecting the target customers is from front to back.

[0103] Through the above implementation, the target quantity can be quickly determined, and then the target customers can be quickly selected using the target quantity.

[0104] S14, extracting a feature matrix from the customer representation matrix of the target customer based on each preset dimension.

[0105] In at least one embodiment of the present invention, the characteristic matrix refers to matrix information that does not include any preset dimension.

[0106] In at least one embodiment of the present invention, the electronic device extracting a feature matrix from the customer characterization matrix of the target customer based on each preset dimension comprises:

[0107] For any preset dimension, the dimension vector corresponding to any preset dimension in the customer representation matrix of the target customer is deleted to obtain the feature matrix.

[0108] By deleting the dimension vector corresponding to any one of the preset dimensions from the customer characterization matrix of the target customer, since there is no need to reconstruct the matrix corresponding to the remaining preset dimensions, the generation efficiency of the feature matrix can be improved.

[0109] S15, generating the influence of each preset dimension on the target customer according to the predicted probability of each feature matrix on the user category and the category probability of the category prediction model.

[0110] In at least one embodiment of the present invention, the predicted probability refers to the probability output by the category prediction model after predicting the feature matrix without any preset dimension on the user category.

[0111] The influence refers to the contribution value of each preset dimension to the target customer in the user category.

[0112] In at least one embodiment of the present invention, the electronic device generates the influence of each preset dimension on the target customer according to the predicted probability of each feature matrix on the user category and the category probability of the category prediction model, including:

[0113] Input each feature matrix into the category prediction model, and extract the output value of each feature matrix on the user category from the normalization layer as the prediction probability;

[0114] The difference between each predicted probability and the category probability is calculated to obtain the influence degree.

[0115] Through the above implementation, it is possible to combine the category prediction model's prediction analysis of the feature matrix that does not have any preset dimension on the user category, and the category prediction model's prediction analysis of the customer representation matrix that has any preset dimension on the user category. Since the influence is analyzed from the same user category, it is possible to accurately determine the influence of any preset dimension on the target customer in the user category.

[0116] S16: Generate characteristic information of the target user according to the influence and the user category.

[0117] In at least one embodiment of the present invention, the feature information includes the user category and the maximum influencing factor of the target customer on the user category, wherein the maximum influencing factor refers to the dimension information of the target user on the preset dimension with the greatest influence.

[0118] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned characteristic information, the above-mentioned characteristic information can also be stored in a node of a blockchain.

[0119] In at least one embodiment of the present invention, the electronic device generating the characteristic information of the target user according to the influence and the user category includes:

[0120] Determine the preset dimension with the greatest influence as the target dimension;

[0121] Extracting information of the target user on the target dimension from the dimension information as a target factor;

[0122] The characteristic information is determined according to the target factor and the user category.

[0123] Through the above implementation, the target factor of the target user in the user category can be accurately located according to the influence degree.

[0124] It can be seen from the above technical solutions that the present invention constructs the customer characterization matrix through a dimension mapping table of each preset dimension. Since each dimension mapping table only stores mapping information of the corresponding preset dimension, the construction efficiency of the customer characterization matrix can be improved. The customer characterization matrix is ​​predicted and analyzed based on the category prediction model. Since the customer characterization matrix combines the dimension information on the multiple preset dimensions, the prediction accuracy of the user category and the category probability can be improved. Then, the target customers can be screened according to the category probability and the user category, and customers with analytical significance can be accurately selected, thereby improving the success rate of business personnel in business development. Furthermore, based on the multiple preset dimensions, the feature matrix can be quickly extracted from the customer characterization matrix of the target customer, thereby improving the generation efficiency of the feature information. In addition, combined with the predicted probability of the feature matrix on the user category and the category probability of the customer characterization matrix of the target customer on the user category, the influence of each preset dimension on the target customer can be accurately determined, thereby improving the positioning accuracy of the feature information.

[0125] like Figure 2 , which is a functional module diagram of a preferred embodiment of the customer feature location device of the present invention. The customer feature location device 11 includes an acquisition unit 110, a generation unit 111, a prediction unit 112, a screening unit 113 and an extraction unit 114. The module / unit referred to in the present invention refers to a series of computer-readable instruction segments that can be acquired by the processor 13 and can complete a fixed function, which is stored in the memory 12. In this embodiment, the functions of each module / unit will be described in detail in subsequent embodiments.

[0126] The acquisition unit 110 acquires dimension information of a preset customer in multiple preset dimensions.

[0127] In at least one embodiment of the present invention, the preset customer refers to a purchasing user of a certain product.

[0128] The multiple preset dimensions include, but are not limited to: customer basic information category, customer transaction behavior category, customer contact behavior category, etc.

[0129] The dimension information refers to the relevant information of the preset customer on the multiple preset dimensions. The dimension information can be obtained from a customer relationship management system.

[0130] The generating unit 111 generates a customer representation matrix of the preset customer according to the dimension mapping table of each preset dimension and the dimension information.

[0131] In at least one embodiment of the present invention, the dimension mapping table stores a mapping relationship between data information on a corresponding preset dimension and a vector. The vector dimension in the dimension mapping table can be 100 dimensions or 300 dimensions. The present invention does not limit the vector dimension of the dimension mapping table.

[0132] The customer characterization matrix is ​​used to indicate the dimension information of the preset customer on the multiple preset dimensions.

[0133] In at least one embodiment of the present invention, the generating unit 111 generates the customer representation matrix of the preset customer according to the dimension mapping table of each preset dimension and the dimension information, including:

[0134] Acquire a list corresponding to each preset dimension from a preset mapping library to obtain a plurality of dimension mapping tables;

[0135] Based on each dimension mapping table, corresponding dimension information is mapped in parallel to obtain a dimension vector of each preset dimension;

[0136] Obtaining a weight matrix of any configuration category in the category prediction model;

[0137] Identify the weight position of the dimension weight of each preset dimension in the weight matrix;

[0138] The dimensional vectors are concatenated according to the weight positions to obtain the customer representation matrix.

[0139] The preset mapping library stores mapping lists of multiple dimensions, and each mapping list stores a mapping relationship between data information and vectors in a corresponding dimension.

[0140] The weight matrix is ​​a matrix generated when training the category prediction model, and the weight matrix is ​​used to quantify the importance of the corresponding preset dimension to any of the configuration categories. It is understandable that in the category prediction model, the weight matrix corresponding to each configuration category is different, and the importance of the preset dimensions to the configuration category in each weight matrix is ​​spliced ​​in the same way. For example, the first row vector in the weight matrix is ​​the importance of the first preset dimension to the configuration category, the second row vector in the weight matrix is ​​the importance of the second preset dimension to the configuration category, and so on.

[0141] By mapping the dimensional information in parallel through the multiple dimensional mapping tables, the generation efficiency of the dimensional vector is improved, and then the dimensional vector is spliced ​​based on the weight position of the dimensional weight in the weight matrix, which can ensure the subsequent prediction accuracy of the user category and the category probability, and improve the rationality of the generation of the customer characterization matrix.

[0142] The prediction unit 112 performs prediction analysis on the customer representation matrix based on a pre-trained category prediction model to obtain user categories and category probabilities of the user categories.

[0143] In at least one embodiment of the present invention, the category prediction model is used to predict the specific category of the customer and the probability that the customer is in the specific category.

[0144] The user categories may include multiple ones. For example, the user category of customer A may be easy to add insurance, easy to refer, and difficult to pay.

[0145] The category probability refers to the probability value that the category prediction model predicts that the preset customer is the user category.

[0146] In at least one embodiment of the present invention, the category prediction model includes a normalization layer and an output layer. The prediction unit 112 performs prediction analysis on the customer representation matrix based on the pre-trained category prediction model to obtain user categories and category probabilities of the user categories, including:

[0147] Calculate the product of each weight matrix and the customer characterization matrix to obtain the output score of the preset customer in each configuration category;

[0148] The output scores are standardized based on the standardization layer to obtain the output probability of the preset customer in each configuration category. The calculation formula of the output probability is:

[0149]

[0150] Among them, P i refers to the output probability of the i-th configuration category, x irefers to the output score of the i-th configuration category, μ refers to the average output score of all configuration categories in the category prediction model, and σ refers to the standard deviation of the output scores of all configuration categories;

[0151] Obtaining a preset probability threshold for each configuration category from the output layer;

[0152] An output probability whose value is greater than a corresponding preset probability threshold is determined as the category probability, and a configuration category corresponding to the category probability is determined as the user category.

[0153] The preset probability threshold is generated when training the category prediction model based on multiple training samples.

[0154] By analyzing the customer characterization matrix through the weight matrix, the output score can be accurately determined, and then the user category can be determined based on the preset probability threshold. Since the preset probability threshold is generated when the category prediction model is trained after multiple training samples, the prediction accuracy of the user category can be improved.

[0155] The screening unit 113 screens target customers from the preset customers according to the user categories and the category probabilities.

[0156] In at least one embodiment of the present invention, the target customer refers to a preset customer with the user category, and the category probability of the preset customer in the user category is ranked in the top N or bottom N, where N refers to the number of the target customers.

[0157] In at least one embodiment of the present invention, the screening unit 113 screens target customers from the preset customers according to the user category and the category probability, including:

[0158] Screening out preset customers corresponding to the user category as customers to be tested;

[0159] Counting the number of customers of the customer to be tested, and calculating the product of the number of customers and a preset ratio to obtain a target number;

[0160] Sort the preset customers according to the category probabilities of the customers to be tested in the user categories to obtain a customer list;

[0161] The target customers are selected from the customer list in sequence according to the target quantity.

[0162] The customer to be tested refers to a preset customer with the user category.

[0163] The preset ratio can be set according to demand, for example, the preset ratio can be 5%.

[0164] The arrangement order of the customer list may be arranged in the order from large to small according to the category probability, or may be arranged in the order from small to large according to the category probability.

[0165] The order of selecting the target customers corresponds to the order of arranging the customer list. For example, if the order of arranging the customer list is from large to small, then the order of selecting the target customers is from front to back.

[0166] Through the above implementation, the target quantity can be quickly determined, and then the target customers can be quickly selected using the target quantity.

[0167] The extraction unit 114 extracts a feature matrix from the customer representation matrix of the target customer based on each preset dimension.

[0168] In at least one embodiment of the present invention, the characteristic matrix refers to matrix information that does not include any preset dimension.

[0169] In at least one embodiment of the present invention, the extracting unit 114 extracts a feature matrix from the customer representation matrix of the target customer based on each preset dimension, including:

[0170] For any preset dimension, the dimension vector corresponding to any preset dimension in the customer representation matrix of the target customer is deleted to obtain the feature matrix.

[0171] By deleting the dimension vector corresponding to any one of the preset dimensions from the customer characterization matrix of the target customer, since there is no need to reconstruct the matrix corresponding to the remaining preset dimensions, the generation efficiency of the feature matrix can be improved.

[0172] The generating unit 111 generates the influence of each preset dimension on the target customer according to the predicted probability of each feature matrix on the user category and the category probability of the category prediction model.

[0173] In at least one embodiment of the present invention, the predicted probability refers to the probability output by the category prediction model after predicting the feature matrix without any preset dimension on the user category.

[0174] The influence refers to the contribution value of each preset dimension to the target customer in the user category.

[0175] In at least one embodiment of the present invention, the generating unit 111 generates the influence of each preset dimension on the target customer according to the predicted probability of each feature matrix on the user category and the category probability of the category prediction model, including:

[0176] Input each feature matrix into the category prediction model, and extract the output value of each feature matrix on the user category from the normalization layer as the prediction probability;

[0177] The difference between each predicted probability and the category probability is calculated to obtain the influence degree.

[0178] Through the above implementation, it is possible to combine the category prediction model's prediction analysis of the feature matrix that does not have any preset dimension on the user category, and the category prediction model's prediction analysis of the customer representation matrix that has any preset dimension on the user category. Since the influence is analyzed from the same user category, it is possible to accurately determine the influence of any preset dimension on the target customer in the user category.

[0179] The generating unit 111 generates characteristic information of the target user according to the influence and the user category.

[0180] In at least one embodiment of the present invention, the feature information includes the user category and the maximum influencing factor of the target customer on the user category, wherein the maximum influencing factor refers to the dimension information of the target user on the preset dimension with the greatest influence.

[0181] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned characteristic information, the above-mentioned characteristic information can also be stored in a node of a blockchain.

[0182] In at least one embodiment of the present invention, the generating unit 111 generates the characteristic information of the target user according to the influence and the user category, including:

[0183] Determine the preset dimension with the greatest influence as the target dimension;

[0184] Extracting information of the target user on the target dimension from the dimension information as a target factor;

[0185] The characteristic information is determined according to the target factor and the user category.

[0186] Through the above implementation, the target factor of the target user in the user category can be accurately located according to the influence degree.

[0187] It can be seen from the above technical solutions that the present invention constructs the customer characterization matrix through a dimension mapping table of each preset dimension. Since each dimension mapping table only stores mapping information of the corresponding preset dimension, the construction efficiency of the customer characterization matrix can be improved. The customer characterization matrix is ​​predicted and analyzed based on the category prediction model. Since the customer characterization matrix combines the dimension information on the multiple preset dimensions, the prediction accuracy of the user category and the category probability can be improved. Then, the target customers can be screened according to the category probability and the user category, and customers with analytical significance can be accurately selected, thereby improving the success rate of business personnel in business development. Furthermore, based on the multiple preset dimensions, the feature matrix can be quickly extracted from the customer characterization matrix of the target customer, thereby improving the generation efficiency of the feature information. In addition, combined with the predicted probability of the feature matrix on the user category and the category probability of the customer characterization matrix of the target customer on the user category, the influence of each preset dimension on the target customer can be accurately determined, thereby improving the positioning accuracy of the feature information.

[0188] like Figure 3 FIG. 1 is a schematic diagram of the structure of an electronic device of a preferred embodiment of the method for locating customer characteristics according to the present invention.

[0189] In one embodiment of the present invention, the electronic device 1 includes, but is not limited to, a memory 12, a processor 13, and computer-readable instructions stored in the memory 12 and executable on the processor 13, such as a customer feature location program.

[0190] Those skilled in the art will understand that the schematic diagram is merely an example of the electronic device 1 and does not constitute a limitation on the electronic device 1. The electronic device 1 may include more or fewer components than shown in the diagram, or a combination of certain components, or different components. For example, the electronic device 1 may also include input and output devices, network access devices, buses, etc.

[0191] The processor 13 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor 13 is the computing core and control center of the electronic device 1, and uses various interfaces and lines to connect various parts of the entire electronic device 1, and executes the operating system of the electronic device 1 and various installed applications, program codes, etc.

[0192] Exemplarily, the computer-readable instructions may be divided into one or more modules / units, which are stored in the memory 12 and executed by the processor 13 to complete the present invention. The one or more modules / units may be a series of computer-readable instruction segments capable of completing specific functions, which are used to describe the execution process of the computer-readable instructions in the electronic device 1. For example, the computer-readable instructions may be divided into an acquisition unit 110, a generation unit 111, a prediction unit 112, a screening unit 113, and an extraction unit 114.

[0193] The memory 12 can be used to store the computer-readable instructions and / or modules, and the processor 13 implements various functions of the electronic device 1 by running or executing the computer-readable instructions and / or modules stored in the memory 12, and calling the data stored in the memory 12. The memory 12 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the electronic device, etc. The memory 12 can include non-volatile and volatile memories, such as: a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other storage devices.

[0194] The memory 12 may be an external memory and / or an internal memory of the electronic device 1. Furthermore, the memory 12 may be a memory in a physical form, such as a memory stick, a TF card (Trans-flash Card), and the like.

[0195] If the module / unit integrated in the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also instruct the relevant hardware to complete it through computer-readable instructions. The computer-readable instructions can be stored in a computer-readable storage medium. When the computer-readable instructions are executed by the processor, the steps of the above-mentioned various method embodiments can be implemented.

[0196] The computer-readable instructions include computer-readable instruction codes, which may be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer-readable instruction code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM).

[0197] The blockchain referred to in the present invention is a new application model of computer technologies such as distributed customer feature positioning, point-to-point transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the blockchain underlying platform, platform product service layer, and application service layer.

[0198] Combination Figure 1 The memory 12 in the electronic device 1 stores computer-readable instructions to implement a method for locating customer characteristics, and the processor 13 can execute the computer-readable instructions to implement:

[0199] Obtain dimension information of a preset customer on multiple preset dimensions;

[0200] Generate a customer representation matrix for the preset customer according to the dimension mapping table of each preset dimension and the dimension information;

[0201] Performing prediction analysis on the customer representation matrix based on a pre-trained category prediction model to obtain user categories and category probabilities of the user categories;

[0202] Screening target customers from the preset customers according to the user category and the category probability;

[0203] Extracting a feature matrix from the customer representation matrix of the target customer based on each preset dimension;

[0204] Generate the influence of each preset dimension on the target customer according to the predicted probability of each feature matrix on the user category and the category probability of the category prediction model;

[0205] The characteristic information of the target user is generated according to the influence and the user category.

[0206] Specifically, the specific implementation method of the processor 13 for the above-mentioned computer readable instructions can refer to Figure 1 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0207] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0208] The computer-readable storage medium stores computer-readable instructions, wherein the computer-readable instructions are used to implement the following steps when executed by the processor 13:

[0209] Obtain dimension information of a preset customer on multiple preset dimensions;

[0210] Generate a customer representation matrix for the preset customer according to the dimension mapping table of each preset dimension and the dimension information;

[0211] Performing prediction analysis on the customer representation matrix based on a pre-trained category prediction model to obtain user categories and category probabilities of the user categories;

[0212] Screening target customers from the preset customers according to the user category and the category probability;

[0213] Extracting a feature matrix from the customer representation matrix of the target customer based on each preset dimension;

[0214] Generate the influence of each preset dimension on the target customer according to the predicted probability of each feature matrix on the user category and the category probability of the category prediction model;

[0215] The characteristic information of the target user is generated according to the influence and the user category.

[0216] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0217] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0218] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0219] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units or devices mentioned can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

[0220] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A customer feature positioning method, characterized in that: The customer feature positioning method comprises: Obtain dimension information of a preset customer on multiple preset dimensions; Generate a customer representation matrix for the preset customer according to the dimension mapping table of each preset dimension and the dimension information; Performing prediction analysis on the customer representation matrix based on a pre-trained category prediction model to obtain user categories and category probabilities of the user categories; Screening target customers from the preset customers according to the user category and the category probability; Extracting a feature matrix from the customer representation matrix of the target customer based on each preset dimension; The influence degree of each preset dimension on the target customer is generated according to the predicted probability of each feature matrix on the user category and the category probability of the category prediction model, including: inputting each feature matrix into the category prediction model, and extracting the output value of each feature matrix on the user category from the normalization layer of the category prediction model as the predicted probability; calculating the difference between each predicted probability and the category probability to obtain the influence degree; Generate characteristic information of the target user according to the influence and the user category.

2. The customer feature positioning method according to claim 1, characterized in that: The step of generating the customer representation matrix of the preset customer according to the dimension mapping table of each preset dimension and the dimension information includes: Acquire a list corresponding to each preset dimension from a preset mapping library to obtain a plurality of dimension mapping tables; Based on each dimension mapping table, corresponding dimension information is mapped in parallel to obtain a dimension vector of each preset dimension; Obtaining a weight matrix of any configuration category in the category prediction model; Identify the weight position of the dimension weight of each preset dimension in the weight matrix; The dimensional vectors are concatenated according to the weight positions to obtain the customer representation matrix.

3. The customer feature positioning method according to claim 2, characterized in that: The category prediction model includes an output layer, and the customer representation matrix is ​​predicted and analyzed based on the pre-trained category prediction model to obtain user categories and category probabilities of the user categories, including: Calculate the product of each weight matrix and the customer characterization matrix to obtain the output score of the preset customer in each configuration category; The output scores are standardized based on the standardization layer to obtain the output probability of the preset customer in each configuration category. The calculation formula of the output probability is: ； in, It refers to The output probability of each configuration category is refers to the The output scores of the configuration categories, It refers to the average of the output scores of all configured categories in the category prediction model. is the standard deviation of the output scores for all configuration categories; Obtaining a preset probability threshold for each configuration category from the output layer; An output probability whose value is greater than a corresponding preset probability threshold is determined as the category probability, and a configuration category corresponding to the category probability is determined as the user category.

4. The method for locating customer characteristics according to claim 2, characterized in that: The extracting of a feature matrix from the customer representation matrix of the target customer based on each preset dimension comprises: For any preset dimension, the dimension vector corresponding to any preset dimension in the customer representation matrix of the target customer is deleted to obtain the feature matrix.

5. The method for locating customer characteristics according to claim 1, characterized in that: The screening of target customers from the preset customers according to the user category and the category probability includes: Screening out preset customers corresponding to the user category as customers to be tested; Counting the number of customers of the customer to be tested, and calculating the product of the number of customers and a preset ratio to obtain a target number; Sort the preset customers according to the category probabilities of the customers to be tested in the user categories to obtain a customer list; The target customers are selected from the customer list in sequence according to the target quantity.

6. The method for locating customer characteristics according to claim 1, characterized in that: The generating of characteristic information of the target user according to the influence and the user category includes: Determine the preset dimension with the greatest influence as the target dimension; Extracting information of the target user on the target dimension from the dimension information as a target factor; The characteristic information is determined according to the target factor and the user category.

7. A customer feature positioning device, characterized in that: The customer feature positioning device comprises: An acquisition unit, used for acquiring dimension information of a preset customer on multiple preset dimensions; A generating unit, configured to generate a customer representation matrix of the preset customer according to a dimension mapping table of each preset dimension and the dimension information; A prediction unit, configured to perform prediction analysis on the customer representation matrix based on a pre-trained category prediction model to obtain user categories and category probabilities of the user categories; A screening unit, used for screening target customers from the preset customers according to the user category and the category probability; An extraction unit, configured to extract a feature matrix from a customer representation matrix of the target customer based on each preset dimension; The generating unit is further used to generate the influence of each preset dimension on the target customer according to the predicted probability of each feature matrix on the user category and the category probability of the category prediction model, including: inputting each feature matrix into the category prediction model, and extracting the output value of each feature matrix on the user category from the normalization layer of the category prediction model as the predicted probability; calculating the difference between each predicted probability and the category probability to obtain the influence; The generating unit is further configured to generate characteristic information of a target user according to the influence and the user category.

8. An electronic device, characterized in that: The electronic device comprises: a memory storing computer-readable instructions; and A processor executes computer-readable instructions stored in the memory to implement the customer feature positioning method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions are executed by a processor in an electronic device to implement the customer feature positioning method as described in any one of claims 1 to 6.

Images