A potential customer analysis and mining method and system

Abstract

The present invention relates to the technical field of data mining, and discloses a potential customer analysis and mining method and system, the method comprising obtaining user-related information and selecting an initial model; constructing a user behavior matrix according to the user-related information; extracting user behavior features according to the user behavior matrix; constructing a training set and a test set according to the user behavior features; inputting the training set into the initial model for training to obtain a user prediction model; using the test set to test the user prediction model, calculating the model accuracy, and deciding to retrain the model according to the model accuracy, or outputting and saving the user prediction model that reaches a certain model accuracy. The method has the following effects: the generated user prediction model can judge potential customers from a customer group with high accuracy and high efficiency according to the customer's behavior, and can adapt to a variety of sales scenarios, thereby improving the generalization performance of the model.

Description

Technical Field

[0001] The present invention relates to the technical field of data mining, and in particular to a potential customer analysis and mining method and system. Background Art

[0002] Potential customer mining refers to finding high-potential customers from various corporate databases and providing them to the corresponding companies for commercial development. As a key part of developing new customers for enterprises, improving its efficiency has important commercial value. Due to the development of data mining technology and methods, this problem has gradually transformed from statistical analysis based on customer data to machine learning based on models. At present, most algorithms in the field of machine learning have studied this problem.

[0003] However, in the current customer mining algorithms, many algorithms do not construct customer features from a comprehensive perspective, but only focus on some dimensions, resulting in insufficient model generalization ability. In addition, algorithms for mining customer information are mostly used for information retrieval, which is often limited to specific scenarios and has limitations when applied to finding potential customers. For example, the potential customer mining algorithm based on sequential patterns depends on the frequency of customer data mining, resulting in the multiplication of computational complexity and data volume, thus limiting the algorithm's ability to process big data. The potential customer mining algorithm based on graph structure depends on the relationship network given by the enterprise. If the network is complex, the computational efficiency will decrease, so it cannot be flexibly applied to all enterprises.

[0004] In short, potential customer mining depends not only on models, but also on the ability to express customer information. On the one hand, many current algorithms do not combine big data well, resulting in models that are not highly accurate and have low computational efficiency. On the other hand, many current algorithms do not construct customer features from a comprehensive perspective, resulting in models that cannot demonstrate good generalization capabilities, and there are certain limitations in applying them to finding potential customers. Summary of the invention

[0005] The present invention provides a potential customer analysis and mining method and system to solve the problems of low accuracy and efficiency of traditional potential customer mining algorithms.

[0006] In a first aspect, in order to solve the above technical problems, the present invention provides a potential customer analysis and mining method, comprising:

[0007] Obtain user attribute information and user behavior information, and select an initial model and accuracy threshold;

[0008] Constructing a user behavior matrix according to the user behavior information;

[0009] Extracting user behavior features according to the user behavior matrix;

[0010] Constructing a training set and a test set according to the user attribute information and the user behavior characteristics;

[0011] Inputting the training set into the initial model for training to obtain an initial user prediction model;

[0012] Using the test set to test the initial user prediction model, and calculating the model accuracy;

[0013] A judgment is made based on the model accuracy and the accuracy threshold, and if the model accuracy is greater than the accuracy threshold, the initial user prediction model is output as a target user prediction model; if the model accuracy is less than the accuracy threshold, the training set is reconstructed and the initial user prediction model is trained;

[0014] The user is predicted according to the target user prediction model to obtain potential customer data.

[0015] In an optional implementation, the obtaining of user attribute information and user behavior information includes directly reading the user attribute information and the user behavior information from a database; wherein the user attribute information includes the user's age and gender; the user behavior information includes specific behavior categories and specific behavior time; wherein the specific behavior categories include browsing, ordering, collecting, sharing, commenting, adding to a shopping cart, or participating in an event.

[0016] In an optional implementation, constructing a user behavior matrix according to the user behavior information includes:

[0017] Setting a weight for each of the user behavior information; wherein the weight represents the importance of each type of user behavior;

[0018] According to the weights, construct a user behavior matrix;

[0019] The user behavior matrix is ​​represented by the following mathematical formula:

[0020]

[0021] in, represents the user behavior matrix, Indicates OK, Indicates List, Indicates User's The weight of each behavior, Indicates the number of users, Indicates the number of user actions.

[0022] In an optional implementation, extracting user behavior features according to the user behavior matrix includes:

[0023] Mapping the user behavior matrix into the latent space of potential factors of user interest to obtain a latent space vector;

[0024] constructing an information matrix according to the latent space vector and the user attribute information;

[0025] Processing redundancy and noise in the information matrix to obtain a feature matrix;

[0026] The user behavior features are extracted from the feature matrix.

[0027] In an optional implementation, the using the test set to test the initial user prediction model and calculating the model accuracy includes:

[0028] The accuracy of the model is calculated according to the following formula:

[0029]

[0030] in, represents the accuracy of the model, represents the number of potential users in the test set and predicted as potential users, represents the number of non-potential users in the test set and predicted as non-potential users, represents the number of non-potential users in the test set but predicted as potential users, Represents the number of potential users in the test set but predicted as non-potential users.

[0031] In an optional implementation, mapping the user behavior matrix into a latent space of potential factors of user interest to obtain a latent space vector includes:

[0032] The latent space vector is calculated according to the following formula:

[0033]

[0034] in, Indicates user The latent space vector corresponding to each behavior is represents the user behavior matrix, Indicates the user's Behavior vector.

[0035] In an optional implementation, constructing an information matrix according to the latent space vector and the user attribute information includes:

[0036] The information matrix is ​​constructed according to the following formula:

[0037]

[0038]

[0039] in, represents the information matrix, Indicates List, Indicates The user feature vector of the column, Indicates user The latent space vector corresponding to each behavior is Indicates the user's Attribute information, represents the amount of user attribute information, Indicates the set zoom factor.

[0040] In an optional implementation manner, the processing of redundancy and noise in the information matrix to obtain a feature matrix includes:

[0041] The feature matrix is ​​calculated according to the following formula:

[0042]

[0043]

[0044] in, represents the feature matrix, Indicates The user feature vector of the column, Indicates the processed The user feature vector of the column, Indicates The probability of the column user feature vector appearing, Indicated in When the column user feature vector appears, The probability of occurrence of column user feature vector.

[0045] In an optional implementation, extracting the user behavior feature from the feature matrix includes:

[0046] Setting the number of extracted user behavior features from the feature matrix; wherein the number of extracted features does not exceed the number of user attribute information;

[0047] Calculate the information content of each column vector of the feature matrix, and select the extracted number of column vectors therefrom in ascending order as the user behavior features;

[0048] The information volume is calculated according to the following formula:

[0049]

[0050] in, Indicates the processed The user feature vector of the column, represents the information amount of the user feature vector, Indicates the processed The probability of occurrence of column user feature vector.

[0051] In a second aspect, the present invention provides a potential customer analysis and mining system, comprising:

[0052] Information initialization module, used to obtain user attribute information and user behavior information, and select an initial model and accuracy threshold;

[0053] A user behavior matrix construction module, used to construct a user behavior matrix according to the user behavior information;

[0054] A user behavior feature extraction module, used to extract user behavior features according to the user behavior matrix;

[0055] A data set building module, used to build a training set and a test set according to the user attribute information and the user behavior characteristics;

[0056] A model training module, used for inputting the training set into the initial model for training to obtain an initial user prediction model;

[0057] An accuracy calculation module, used to test the initial user prediction model using the test set and calculate the model accuracy;

[0058] A model judgment module, configured to make a judgment based on the model accuracy and the accuracy threshold, and output the initial user prediction model as a target user prediction model if the model accuracy is greater than the accuracy threshold; and reconstruct the training set and train the initial user prediction model if the model accuracy is less than the accuracy threshold;

[0059] The result prediction module is used to predict users according to the target user prediction model to obtain potential customer data.

[0060] In a third aspect, the present invention also provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the potential customer analysis and mining method described in any one of the above items is implemented.

[0061] In a fourth aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the potential customer analysis and mining methods described above.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] The present invention provides a potential customer analysis and mining method, comprising the steps of obtaining user attribute information and user behavior information, and selecting an initial model and an accuracy threshold; constructing a user behavior matrix according to the user behavior information; extracting user behavior features according to the user behavior matrix; constructing a training set and a test set according to the user attribute information and the user behavior features; inputting the training set into the initial model for training to obtain an initial user prediction model; using the test set to test the initial user prediction model to calculate the model accuracy; making a judgment according to the model accuracy and the accuracy threshold, if the model accuracy is greater than the accuracy threshold, outputting the initial user prediction model as a target user prediction model; if the model accuracy is less than the accuracy threshold, reconstructing the training set and training the initial user prediction model; predicting users according to the target user prediction model to obtain potential customer data.

[0064] The method obtains relevant information of customers from a database, processes the relevant information and generates a training set and a test set, and finally substitutes the information into a pre-selected model for training, thereby obtaining a user prediction model. The user prediction model can accurately and efficiently identify potential customers from a customer group based on customer behavior, and can adapt to a variety of sales scenarios, thereby improving the generalization performance of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a flowchart of a potential customer analysis and mining method provided by the first embodiment of the present invention;

[0066] Figure 2 It is a schematic diagram of the structure of a potential customer analysis and mining system provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0067] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0068] Reference Figure 1 The first embodiment of the present invention provides a potential customer analysis and mining method, comprising the following steps:

[0069] S1, obtain user attribute information and user behavior information, and select an initial model and accuracy threshold;

[0070] S2, constructing a user behavior matrix according to the user behavior information;

[0071] S3, extracting user behavior features according to the user behavior matrix;

[0072] S4, constructing a training set and a test set according to the user attribute information and the user behavior characteristics;

[0073] S5, inputting the training set into the initial model for training to obtain an initial user prediction model;

[0074] S6, using the test set to test the initial user prediction model, and calculating the model accuracy;

[0075] S7, judging according to the model accuracy and the accuracy threshold, if the model accuracy is greater than the accuracy threshold, outputting the initial user prediction model as the target user prediction model; if the model accuracy is less than the accuracy threshold, reconstructing the training set and training the initial user prediction model;

[0076] S8, predicting users according to the target user prediction model to obtain potential customer data.

[0077] It should be noted that potential customer discovery, as one of the core strategies for corporate market expansion, aims to deeply explore potential customer groups with high growth potential in corporate databases, and then accurately locate the target objects for corporate business development activities. Although the research on potential customer discovery algorithms is developing rapidly, the existing customer mining algorithms are often one-sided when constructing customer features, focusing only on a few specific dimensions and ignoring the customer's comprehensive information, which directly leads to the limited generalization ability of the model when facing unknown or diverse customers. In addition, most of these algorithms originated from the field of information retrieval, and the difference between their original design intentions and actual application scenarios makes them show certain limitations in the task of potential customer discovery.

[0078] In the present invention, the attribute information and behavior information of the user are first obtained, and then the above information is constructed into a corresponding matrix according to the principles of algorithms and information theory, and the behavior characteristics of the user are extracted from it. Then the behavior characteristics are input into a pre-selected model for training, thereby obtaining a target user prediction model, and finally the target user prediction model is used to predict the customer group, and finally the potential user is obtained. In the process of matrix construction, the present invention introduces latent space representation and information matrix, combines customers and projects from the implicit perspective of potential customers, and converts the two into supervised tasks. Similar to the graph structure, the use of latent space allows end-to-end calculation of potential customers, which makes up for the deficiency of the current algorithm's over-reliance on the relational network. At the same time, with the powerful ability of the neural network model, the mining algorithm can make full use of the huge amount of data. In summary, the present invention improves the efficiency of the accuracy of potential customer mining by proposing a mining algorithm based on latent space and big data, thereby generating greater value for enterprises to mine potential customers.

[0079] In step S1, user attribute information and user behavior information are obtained from the database, and a suitable initial model and accuracy threshold are selected according to the specific application scenario.

[0080] Specifically, the user attribute information and the user behavior information are first directly obtained from a database storing user-related information. Then, according to the specific application scenario, a suitable initial model and an accuracy threshold corresponding to the model are selected. The user attribute information includes the user's age, gender, and income; the user behavior information includes specific behavior categories, specific behavior time, and specific behavior spending amount; the specific behavior categories include browsing, ordering, collecting, sharing, commenting, adding to shopping carts, and participating in activities.

[0081] It should be noted that when selecting a suitable initial model, it is necessary to accurately analyze the requirements of the application scenario. For example, if the amount or dimension of user-related information is small, you can choose algorithms suitable for small samples such as support vector machines and decision trees; on the contrary, if the amount or dimension of user-related information is large, you can choose algorithms suitable for large samples such as convolutional neural networks and recurrent neural networks. At the same time, when selecting the initial model, you can choose to train a model that has never been trained, which can be more suitable for the current application scenario; you can also choose a trained model and then use the training set for learning, which can adapt to better generalization performance.

[0082] In step S2, a user behavior matrix is ​​constructed based on the obtained user behavior information.

[0083] In one embodiment, constructing a user behavior matrix based on the user behavior information includes setting a weight for each of the user behavior information; wherein the weight represents the importance of each type of user behavior; constructing a user behavior matrix based on the weight; wherein the user behavior matrix is ​​represented by the following mathematical formula,

[0084]

[0085] in, represents the user behavior matrix, Indicates OK, Indicates List, Indicates User's The weight of each behavior, Indicates the number of users, Indicates the number of user actions.

[0086] Specifically, first, different weights are set for each user behavior information according to the importance of different behaviors. Then, according to each behavior of each user, the weights of all users and all behaviors are integrated to form the user behavior matrix. Among them, the weights need to be set to different values ​​according to the number and importance of behaviors. For example, there are three behaviors, namely browsing, ordering and collecting, and the weights of the three behaviors can be set to 1, 10 and 5 respectively.

[0087] It should be noted that since there may be large fluctuations between the values ​​of the weights, this will cause the model to be unable to learn the behaviors with smaller weights during subsequent training, thereby causing the model to be underfitted. Therefore, before forming the user behavior matrix, the weights can be normalized. The weights are normalized according to the following formula:

[0088]

[0089] in, represents the weight, represents the normalized weight, represents the minimum value among the weights, Represents the maximum value among the weights.

[0090] In step S3, extracting user behavior features according to the user behavior matrix includes the following steps:

[0091] S31, mapping the user behavior matrix to the latent space of the user's interest potential factors to obtain a latent space vector;

[0092] S32, constructing an information matrix according to the latent space vector and the user attribute information;

[0093] S33, processing redundancy and noise in the information matrix to obtain a feature matrix;

[0094] S34, extracting the user behavior features from the feature matrix.

[0095] In step S31, mapping the user behavior matrix to the latent space of the user interest potential factors to obtain the latent space vector includes calculating the latent space vector according to the following formula:

[0096]

[0097] in, Indicates user The latent space vector corresponding to each behavior is represents the user behavior matrix, Indicates the user's Behavior vector.

[0098] Specifically, we first need to analyze the user's interests, find out the potential factors that affect the user's interests, and determine the dimensions of the latent space and the attributes represented by each dimension based on these potential factors. Then, we use the set formula to map the user behavior matrix to the constructed latent space to obtain the latent space vector. Among them, analyzing the user's interests requires the help of some knowledge of psychology and behavior.

[0099] It should be noted that the latent space of potential factors of user interest refers to the space composed of intrinsic factors that are not easily observed or quantified directly but have a profound impact on user interest, in addition to explicit factors such as user behavior, preferences, and needs. The latent space contains the user's potential interests, unexpressed needs, and complex emotions and behavioral tendencies affected by multiple factors such as situational, psychological, and social factors. Therefore, constructing a good latent space and mapping the user's interests into the latent space can better determine the user's potential interests, thereby improving the accuracy of feature extraction.

[0100] In step S32, constructing an information matrix according to the latent space vector and the user attribute information includes constructing the information matrix according to the following formula:

[0101]

[0102]

[0103] in, represents the information matrix, Indicates List, Indicates The user feature vector of the column, Indicates user The latent space vector corresponding to each behavior is Indicates the user's Attribute information, represents the amount of user attribute information, Indicates the set zoom factor.

[0104] Specifically, the latent space vector and the user attribute information obtained by user interest mapping are acquired, and a user feature vector is calculated according to a formula, and then the user feature vectors are combined in sequence to obtain an information matrix.

[0105] It should be noted that the scaling factor in the formula is essentially used to determine the proportional relationship between the user behavior information and the user attribute information, that is, the importance between the two, and its value is a decimal between 0 and 1. Therefore, when setting the scaling factor, it is necessary to consider the relationship between the user behavior information and the user attribute information. In general, the user's behavior is more important than the user's attributes, so the scaling factor can be set between 0.1 and 0.5.

[0106] In step S33, the processing of the redundancy and noise in the information matrix to obtain a feature matrix includes calculating the feature matrix according to the following formula:

[0107]

[0108]

[0109] in, represents the feature matrix, Indicates The user feature vector of the column, Indicates the processed The user feature vector of the column, Indicates The probability of the column user feature vector appearing, Indicated in When the column user feature vector appears, The probability of occurrence of column user feature vector.

[0110] Specifically, after reading the information matrix, the redundancy and noise in the information matrix can be processed by the conditional information entropy formula based on the principle of conditional information entropy, so as to reduce the redundancy and noise and obtain the feature matrix.

[0111] It should be noted that information entropy represents the degree of confusion or disorder of information, while conditional information entropy refers to the information entropy of another information under the influence of one information. By calculating the conditional information entropy of user feature vectors, the main features of users can be extracted from them according to the relationship between different user feature vectors, and the influence of other features on the main features can be weakened, so that the user's features can be obtained more accurately.

[0112] In step S34, the extracting of the user behavior features from the feature matrix includes setting the number of extractions of the user behavior features from the feature matrix; wherein the extraction number does not exceed the number of the user attribute information; calculating the amount of information of each column vector of the feature matrix, and selecting column vectors whose number is the extraction number in ascending order as the user behavior features; wherein the information amount is calculated according to the following formula,

[0113]

[0114] in, Indicates the processed The user feature vector of the column, represents the information amount of the user feature vector, Indicates the processed The probability of occurrence of column user feature vector.

[0115] Specifically, first, according to the application scenario and the number of the user behavior features, the number of extractions of the user behavior features to be extracted from the feature matrix is ​​set. Then, the amount of information of each column vector of the feature matrix is ​​calculated, and each column vector is sorted from small to large according to the corresponding amount of information. Finally, from the sorted column vectors, the number of column vectors equal to the extraction number is selected from small to large as the user behavior features.

[0116] It should be noted that the amount of information refers to the information measure required to select an event from multiple equally possible events. The larger the value of the amount of information, the smaller the possibility of the event, that is, the less likely it is to occur. Therefore, by calculating the amount of information for each user feature vector and selecting several user feature vectors with relatively small amounts of information, the user's preferred choice can be obtained, thereby improving the prediction accuracy.

[0117] In step S4, the user attribute information and the user behavior characteristics are split according to a certain ratio to construct a training set and a test set.

[0118] Specifically, the user attribute information and the user behavior characteristics are obtained, and the user attribute information and the user behavior characteristics of the same user are stored in correspondence as a data set. Then, a split ratio of the data set is set, and the data set is split according to the ratio, respectively, as a training set and a test set.

[0119] It should be noted that when splitting the data set into training set and test set, it is necessary to ensure the randomness of the samples, and the training set and test set should be independent of each other, that is, the samples in the test set should not be seen or affected by the model during the training process. At the same time, the training set accounts for a large proportion, and the test set accounts for a small proportion. The commonly used ratio of training set to test set is about 7:3; if the number of samples in the data set is large enough, the ratio of the training set can be appropriately increased. In addition, it is necessary to deal with missing values, outliers and noise data, and perform necessary data conversion and normalization to improve the learning efficiency and performance of the model.

[0120] In step S5, the training set is input into the initial model for training, and an initial user prediction model is obtained after training.

[0121] Specifically, firstly, the format of the data in the training set needs to be adjusted according to the initial model to conform to the input of the initial model, and then the training set is input into the initial model for training to obtain the initial user prediction model and save it.

[0122] It should be noted that during the training process, the necessary information of each round of training should be saved in time, such as the loss, gradient value and parameter value of the current round, so that the location of the error can be found in time when an error occurs, or more appropriate initial parameters can be selected based on the information.

[0123] In step S6, the obtained initial user prediction model is tested using the test set, and the model accuracy is calculated according to a formula.

[0124] In one embodiment, the using the test set to test the initial user prediction model to calculate the model accuracy includes calculating the model accuracy according to the following formula:

[0125]

[0126] in, represents the accuracy of the model, represents the number of potential users in the test set and predicted as potential users, represents the number of non-potential users in the test set and predicted as non-potential users, represents the number of non-potential users in the test set but predicted as potential users, Represents the number of potential users in the test set but predicted as non-potential users.

[0127] Specifically, the trained initial user prediction model is first tested using the test set to obtain the prediction result for each user, and then the accuracy of the model is calculated based on the existing results of the test set and the results of the prediction using the model.

[0128] Furthermore, in order to more accurately determine the effect of the initial user prediction model, the model precision and model recall can also be calculated. The model precision is calculated according to the following formula:

[0129]

[0130] in, represents the precision of the model, represents the number of potential users in the test set and predicted as potential users, It represents the number of non-potential users in the test set but predicted to be potential users.

[0131] The recall rate of the model is calculated according to the following formula:

[0132]

[0133] in, represents the recall rate of the model, represents the number of potential users in the test set and predicted as potential users, Represents the number of potential users in the test set but predicted as non-potential users.

[0134] It should be noted that the model precision and the model recall are inversely related, that is, increasing one will lead to a decrease in the other. Therefore, in practical applications, it is necessary to balance the model precision and the model recall according to specific needs.

[0135] In step S7, the size of the model accuracy and the accuracy threshold is determined, and corresponding operations are performed according to the size results.

[0136] In one implementation, if the model accuracy is greater than the accuracy threshold, the initial user prediction model is directly output as a target user prediction model, and the structure and parameters of the target user prediction model are saved in a file.

[0137] Furthermore, some models have many structures and parameters, and directly saving them will take up a lot of storage space. Therefore, a model compression algorithm can be used to simplify the structure of the model, or convert the model parameters from floating point type to integer type to compress the size of the model.

[0138] In another embodiment, if the model accuracy is less than the accuracy threshold, the training set needs to be reconstructed and trained again based on the initial user prediction model. Reconstructing the training set requires obtaining more training data, and the method used can be to obtain additional data from a database, or to randomly generate data within a certain range based on the original data in the training set.

[0139] In step S8, the user group data to be predicted is first obtained, and then the target user prediction model is used to predict the user group to obtain potential customer data.

[0140] To facilitate understanding of the present invention, the working process of the present invention is described below using a relatively common scenario as an example.

[0141] In this embodiment, it is planned to use a potential customer analysis and mining system to analyze and obtain potential customers in a customer group. This system is intended to accurately and quickly find potential customers from a specific customer group.

[0142] Step 1: Obtain user attribute information and user behavior information from the established database, and select a suitable initial model and the corresponding accuracy threshold.

[0143] Step 2: Establish a user behavior matrix based on user behavior information.

[0144] Step 3: Extract the user behavior features from the obtained user behavior matrix.

[0145] Step 4: Create training sets and test sets of appropriate sizes based on the obtained user behavior characteristics and user attribute information.

[0146] Step 5: Input the training set into the initial model for training, obtain the initial user prediction model, and store it.

[0147] Step 6: Input the test set into the initial user prediction model for testing, and calculate the model accuracy based on the test results.

[0148] Step 7: Determine the size of the model accuracy and the accuracy threshold; if the model accuracy is greater than the accuracy threshold, then the initial user prediction model is output as the target user prediction model and saved; if the model accuracy is less than the accuracy threshold, it is necessary to rebuild the training set and train the initial user prediction model again until the model accuracy is greater than the accuracy threshold.

[0149] Step 8: Find the corresponding customer group and use the target user prediction model to predict the customer group to obtain potential customer data.

[0150] Through the above steps, the system can use the target user prediction model to predict the customer group and obtain potential customer data accurately and quickly.

[0151] In summary, the present invention provides a potential customer analysis and mining method, including obtaining user attribute information and user behavior information, and selecting an initial model and an accuracy threshold; constructing a user behavior matrix according to the user behavior information; extracting user behavior features according to the user behavior matrix; constructing a training set and a test set according to the user attribute information and the user behavior features; inputting the training set into the initial model for training to obtain an initial user prediction model; using the test set to test the initial user prediction model to calculate the model accuracy; judging according to the model accuracy and the accuracy threshold, if the model accuracy is greater than the accuracy threshold, outputting the initial user prediction model as a target user prediction model; if the model accuracy is less than the accuracy threshold, reconstructing the training set and training the initial user prediction model; predicting the user according to the target user prediction model to obtain potential customer data. The method obtains relevant information of the customer from the database, then processes the relevant information and generates a training set and a test set, and finally substitutes them into the pre-selected model for training, thereby obtaining a user prediction model. The user prediction model can identify potential customers from a customer group based on customer behavior with high accuracy and efficiency, and can adapt to a variety of sales scenarios, thereby improving the generalization performance of the model.

[0152] Reference Figure 2 The second embodiment of the present invention provides a potential customer analysis and mining system, including:

[0153] Information initialization module, used to obtain user attribute information and user behavior information, and select an initial model and accuracy threshold;

[0154] A user behavior matrix construction module, used to construct a user behavior matrix according to the user behavior information;

[0155] The user behavior matrix construction module includes a weight setting module for setting a weight for each of the user behavior information; and a matrix construction module for constructing the user behavior matrix according to the weight.

[0156] A user behavior feature extraction module, used to extract user behavior features according to the user behavior matrix;

[0157] Among them, the user behavior feature extraction module includes a vector mapping module, which is used to map the user behavior matrix into the latent space to obtain the latent space vector; an information matrix construction module, which is used to construct the information matrix according to the latent space vector; a feature matrix construction module, which is used to process the information matrix to obtain the feature matrix; and a feature extraction module, which is used to extract the behavior features of each user from the feature matrix.

[0158] A data set building module, used to build a training set and a test set according to the user attribute information and the user behavior characteristics;

[0159] Among them, the data set establishment module includes a data processing module, which is used to process missing values, abnormal values ​​and noise data in the user attribute information and the user behavior characteristics; and a data set splitting module, which is used to split the data set into the training set and the test set according to a specific ratio.

[0160] A model training module, used for inputting the training set into the initial model for training to obtain an initial user prediction model;

[0161] An accuracy calculation module, used to test the initial user prediction model using the test set and calculate the model accuracy;

[0162] A model judgment module, configured to make a judgment based on the model accuracy and the accuracy threshold, and output the initial user prediction model as a target user prediction model if the model accuracy is greater than the accuracy threshold; and reconstruct the training set and train the initial user prediction model if the model accuracy is less than the accuracy threshold;

[0163] The result prediction module is used to predict users according to the target user prediction model to obtain potential customer data.

[0164] It should be noted that a potential customer analysis and mining system provided in an embodiment of the present invention is used to execute all process steps of a potential customer analysis and mining method in the above embodiment, and the working principles and beneficial effects of the two correspond one to one, so they will not be repeated here.

[0165] The embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a potential customer mining program based on big data analysis. When the processor executes the computer program, the steps in the above-mentioned various potential customer analysis and mining method embodiments are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of each module / unit in the above-mentioned system embodiments are realized, such as the accuracy calculation module.

[0166] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0167] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than the above components, or may combine certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0168] The processor may be a central processing unit (CPU), 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 is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.

[0169] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory 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 mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media 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 volatile solid-state storage devices.

[0170] Wherein, if the module / unit integrated in the electronic device 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 such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0171] It should be noted that the system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the drawings of the system embodiment provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0172] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A potential customer analysis and mining method, characterized in that: include: Obtain user attribute information and user behavior information, and select an initial model and accuracy threshold; Constructing a user behavior matrix according to the user behavior information; Extracting user behavior features according to the user behavior matrix; Constructing a training set and a test set according to the user attribute information and the user behavior characteristics; Inputting the training set into the initial model for training to obtain an initial user prediction model; Using the test set to test the initial user prediction model, and calculating the model accuracy; A judgment is made based on the model accuracy and the accuracy threshold, and if the model accuracy is greater than the accuracy threshold, the initial user prediction model is output as a target user prediction model; if the model accuracy is less than the accuracy threshold, the training set is reconstructed and the initial user prediction model is trained; Predicting users according to the target user prediction model to obtain potential customer data; The step of extracting user behavior features according to the user behavior matrix includes: Mapping the user behavior matrix into the latent space of potential factors of user interest to obtain a latent space vector; constructing an information matrix according to the latent space vector and the user attribute information; Processing redundancy and noise in the information matrix to obtain a feature matrix; Extracting the user behavior feature from the feature matrix; The step of mapping the user behavior matrix to the latent space of potential factors of user interest to obtain a latent space vector includes: The latent space vector is calculated according to the following formula: ;in, Indicates user The latent space vector corresponding to each behavior is represents the user behavior matrix, Indicates the user's Behavior vectors; The step of constructing an information matrix according to the latent space vector and the user attribute information includes: The information matrix is ​​constructed according to the following formula: , ;in, represents the information matrix, Indicates List, Indicates The user feature vector of the column, Indicates user The latent space vector corresponding to each behavior is Indicates the user's Attribute information, represents the amount of user attribute information, Indicates the set scaling factor; The processing of the redundancy and noise in the information matrix to obtain a feature matrix includes: The feature matrix is ​​calculated according to the following formula: , ;in, represents the feature matrix, Indicates The user feature vector of the column, Indicates the processed The user feature vector of the column, Indicates The probability of the column user feature vector appearing, Indicated in When the column user feature vector appears, The probability of the column user feature vector appearing.

2. The potential customer analysis and mining method according to claim 1, characterized in that: The obtaining of user attribute information and user behavior information includes directly reading the user attribute information and the user behavior information from a database; wherein the user attribute information includes the user's age and gender; the user behavior information includes specific behavior categories and specific behavior time; wherein the specific behavior categories include browsing, ordering, collecting, sharing, commenting, adding to a shopping cart, or participating in an event.

3. The potential customer analysis and mining method according to claim 1, characterized in that: The step of constructing a user behavior matrix according to the user behavior information includes: Setting a weight for each of the user behavior information; wherein the weight represents the importance of each type of user behavior; According to the weights, construct a user behavior matrix; The user behavior matrix is ​​represented by the following mathematical formula: ;in, represents the user behavior matrix, Indicates OK, Indicates List, Indicates User's The weight of each behavior, Indicates the number of users, Indicates the number of user actions.

4. The potential customer analysis and mining method according to claim 1, characterized in that: The using the test set to test the initial user prediction model and calculating the model accuracy includes: The accuracy of the model is calculated according to the following formula: ;in, represents the accuracy of the model, represents the number of potential users in the test set and predicted as potential users, represents the number of non-potential users in the test set and predicted as non-potential users, represents the number of non-potential users in the test set but predicted as potential users, Represents the number of potential users in the test set but predicted as non-potential users.

5. The potential customer analysis and mining method according to claim 1, characterized in that: The extracting the user behavior feature from the feature matrix includes: Setting the number of extracted user behavior features from the feature matrix; wherein the number of extracted features does not exceed the number of user attribute information; Calculate the information content of each column vector of the feature matrix, and select the column vectors whose number is the extracted number from them in ascending order as the user behavior features; The information volume is calculated according to the following formula: ;in, Indicates the processed The user feature vector of the column, represents the information amount of the user feature vector, Indicates the processed The probability of the column user feature vector appearing.

6. A potential customer analysis and mining system, characterized in that: The method for analyzing and mining potential customers as claimed in any one of claims 1 to 5 comprises: Information initialization module, used to obtain user attribute information and user behavior information, and select an initial model and accuracy threshold; A user behavior matrix construction module, used to construct a user behavior matrix according to the user behavior information; A user behavior feature extraction module, used to extract user behavior features according to the user behavior matrix; A data set building module, used to build a training set and a test set according to the user attribute information and the user behavior characteristics; A model training module, used for inputting the training set into the initial model for training to obtain an initial user prediction model; An accuracy calculation module, used to test the initial user prediction model using the test set and calculate the model accuracy; A model judgment module, configured to make a judgment based on the model accuracy and the accuracy threshold, and output the initial user prediction model as a target user prediction model if the model accuracy is greater than the accuracy threshold; and reconstruct the training set and train the initial user prediction model if the model accuracy is less than the accuracy threshold; A result prediction module, used to predict users according to the target user prediction model to obtain potential customer data; The step of extracting user behavior features according to the user behavior matrix includes: Mapping the user behavior matrix into the latent space of potential factors of user interest to obtain a latent space vector; constructing an information matrix according to the latent space vector and the user attribute information; Processing redundancy and noise in the information matrix to obtain a feature matrix; Extracting the user behavior feature from the feature matrix; The step of mapping the user behavior matrix to the latent space of potential factors of user interest to obtain a latent space vector includes: The latent space vector is calculated according to the following formula: ;in, Indicates user The latent space vector corresponding to each behavior is represents the user behavior matrix, Indicates the user's Behavior vectors; The step of constructing an information matrix according to the latent space vector and the user attribute information includes: The information matrix is ​​constructed according to the following formula: , ;in, represents the information matrix, Indicates List, Indicates The user feature vector of the column, Indicates user The latent space vector corresponding to each behavior is Indicates the user's Attribute information, represents the amount of user attribute information, Indicates the set scaling factor; The processing of the redundancy and noise in the information matrix to obtain a feature matrix includes: The feature matrix is ​​calculated according to the following formula: , ;in, represents the feature matrix, Indicates The user feature vector of the column, Indicates the processed The user feature vector of the column, Indicates The probability of the column user feature vector appearing, Indicated in When the column user feature vector appears, The probability of the column user feature vector appearing.

Images