A potential label mining method and device

Abstract

The embodiment of the present invention provides a potential tag mining method and device, which can be used in the field of big data technology. The method includes: preprocessing the original data of the user to be mined to generate user location data; generating high-frequency location features based on the user location data; matching the high-frequency location features with the target location features of the target user obtained through a tag mining model to generate potential tags for the user to be mined, accurately locating the user type, and assigning accurate weights to each indicator, thereby improving the accuracy of the tag mining model and formulating effective marketing strategies.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, in particular to the field of big data technology, and more particularly to a potential tag mining method and device. Background Art

[0002] At present, traditional enterprises and many e-commerce companies usually use user segmentation models (RFM models) to classify and predict users, so as to formulate different marketing strategies for users of different categories. The traditional RFM model scores users based on the segmentation of each indicator according to experience, without a unified and scientific scoring standard, resulting in poor accuracy in user classification and positioning. In addition, there are multicollinearity problems in the consumption frequency and consumption amount in the RFM model, resulting in inaccurate weights of each indicator, which leads to low accuracy of the RFM model and inability to formulate effective marketing strategies. Summary of the invention

[0003] One object of the present invention is to provide a potential tag mining method that can accurately locate user types and assign accurate weights to each indicator, thereby improving the accuracy of the tag mining model and formulating effective marketing strategies. Another object of the present invention is to provide a potential tag mining device. Another object of the present invention is to provide a computer-readable medium. Another object of the present invention is to provide a computer device.

[0004] In order to achieve the above objectives, the present invention discloses a potential tag mining method, comprising:

[0005] Preprocess the original data of the users to be mined to generate user location data;

[0006] Generate high-frequency location features based on user location data;

[0007] The tag mining model is used to match the high-frequency location features with the acquired target location features of the target user to generate potential tags for the user to be mined.

[0008] Preferably, preprocessing the original data of the user to be mined to generate user location data includes:

[0009] According to the user identification code of the user to be mined, the original data is divided to generate the original data corresponding to the user identification code;

[0010] The original data corresponding to the user identification code is cleaned to generate user location data.

[0011] Preferably, the user location data includes time and location coordinates; the high-frequency location features include high-frequency locations to be mined;

[0012] Generate high-frequency location features based on user location data, including:

[0013] Statistics are performed on the time and location coordinates according to a preset period to generate the location frequency of the user to be mined at each location coordinate;

[0014] Sort the position coordinates by position frequency;

[0015] Select the high-frequency position to be mined at the first specified sorting position.

[0016] Preferably, the high-frequency location features include high-frequency administrative codes to be mined;

[0017] Generate high-frequency location features based on user location data, including:

[0018] Query the administrative code corresponding to the location coordinates;

[0019] Count the number of each administrative code and sort the administrative codes according to the counted number;

[0020] Select the high-frequency administrative code to be mined in the second specified sorting position.

[0021] Preferably, the high-frequency position feature includes the difference feature to be mined; the target position feature includes the target highest-frequency position;

[0022] Generate high-frequency location features based on user location data, including:

[0023] Select the highest frequency position to be mined from the sorted position coordinates;

[0024] Through the weighted Euclidean distance formula, the difference features to be mined are generated according to the target highest frequency position and the highest frequency position to be mined.

[0025] Preferably, before matching the high-frequency location features with the acquired target location features of the target user through the tag mining model to generate potential tags of the user to be mined, the method further includes:

[0026] By specifying a classifier, the high-frequency position features are evaluated to generate feature weights corresponding to each dimension of high-frequency position features;

[0027] The feature weights are sorted, and the high-frequency position features corresponding to the feature weights of the third specified sorting positions are selected.

[0028] Preferably, after sorting the feature weights and selecting the high-frequency position feature corresponding to the feature weight of the first specified sorting position, the method further includes:

[0029] By specifying the dimensionality reduction method, multiple high-frequency position features are transformed into low-dimensional space to generate key indicator features after dimensionality reduction.

[0030] Preferably, before matching the high-frequency location features with the acquired target location features of the target user through the tag mining model to generate potential tags of the user to be mined, the method further includes:

[0031] Obtain target location features of target users, including target high-frequency locations, target high-frequency administrative codes, target difference features, and training labels;

[0032] Through supervised learning methods, the target high-frequency positions, target high-frequency administrative codes, target difference features and training labels are trained to build a label mining model.

[0033] The present invention also discloses a potential tag mining device, comprising:

[0034] A preprocessing unit, used to preprocess the original data of the user to be mined to generate user location data;

[0035] A generating unit, used for generating high-frequency location features according to user location data;

[0036] The matching unit is used to match the high-frequency position features with the acquired target position features of the target user through the label mining model to generate potential labels of the user to be mined.

[0037] The present invention also discloses a computer-readable medium on which a computer program is stored. When the program is executed by a processor, the method described above is implemented.

[0038] The present invention also discloses a computer device, including a memory and a processor, wherein the memory is used to store information including program instructions, the processor is used to control the execution of program instructions, and the processor implements the above method when executing the program.

[0039] The present invention pre-processes the original data of the user to be mined to generate user location data; generates high-frequency location features based on the user location data; matches the high-frequency location features with the target location features of the target user obtained through a label mining model to generate potential labels for the user to be mined, can accurately locate the user type, and assign accurate weights to each indicator, thereby improving the accuracy of the label mining model and formulating effective marketing strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A schematic diagram of the structure of a traditional RFM model provided by an embodiment of the present invention;

[0042] Figure 2 A flowchart of a potential tag mining method provided by an embodiment of the present invention;

[0043] Figure 3 A flowchart of another potential tag mining method provided by an embodiment of the present invention;

[0044] Figure 4 A schematic diagram of the structure of a potential tag mining device provided by an embodiment of the present invention;

[0045] Figure 5 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] 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.

[0047] It should be noted that the potential label mining method and device disclosed in the present application can be used in the field of big data technology, and can also be used in any field other than the field of big data technology. The application field of the potential label mining method and device disclosed in the present application is not limited.

[0048] In order to facilitate the understanding of the technical solution provided by the present application, the relevant contents of the technical solution of the present application are first explained below. User tags are user information tags, that is, to classify and locate users, and to describe the characteristics and interests of users through vocabulary tags with strong generalization. User tags are the core factors of user portraits. By comprehensively analyzing the behavioral data generated by users in the platform, user tags with differential characteristics are generated. User tags are visual information for understanding user behavior, preferences and habits, and provide important information and data basis for user portraits. User tags include static tags, dynamic tags and predictive tags, among which, static tags can be used to judge user basic needs, dynamic tags can be used to improve user experience, and predictive tags can be used to improve user conversion rate and product value. Static tags are obtained through the basic information of users, and the basic information is fixed data of users, including but not limited to name, gender and age. Dynamic tags are obtained through the user's operation behavior, and the operation behavior includes all operations from starting the platform to closing the platform. Predictive tags are obtained by predicting future behaviors and preferences based on the user's operation behavior, and through the user's predicted tags, personalized marketing strategies can be formulated for users to achieve accurate push, thereby improving user conversion rate.

[0049] In order to achieve accurate push notifications to users and improve user conversion rates, high requirements are placed on the accuracy and timeliness of predicting users' potential tags. In related technologies, it is proposed to use the traditional RFM model to classify users, that is, to predict users' potential tags. Figure 1 A schematic diagram of the structure of a traditional RFM model provided by an embodiment of the present invention is shown in FIG. Figure 1As shown in the figure, R stands for Recency, which indicates the user's most recent consumption time, F stands for Frequency, which indicates the number of purchases made by the user within a specified time, i.e., consumption frequency, and M stands for Monetary, which indicates the user's consumption amount. First, user data is collected. The user data includes the user identification code (ID), purchase time, and purchase amount. The most recent consumption time, consumption frequency, and consumption amount of each user within a specified time are calculated through the user data; the user's potential label is predicted based on the influencing factors corresponding to the user's most recent consumption time, consumption frequency, and consumption amount within a specified time. Specifically, the influencing factors corresponding to the most recent consumption time include store memory strength, number of contact opportunities, and repurchase cycle; the influencing factors corresponding to the consumption frequency include brand loyalty, store familiarity, customer membership level, and purchase habit formation; the influencing factors corresponding to the consumption amount include consumption ability and product recognition; and the user's potential label is predicted based on the influencing factors corresponding to the user data according to experience. The traditional RFM model defines users with a recent consumption time, high consumption frequency, and high consumption amount as high-quality users. The prediction of the user's potential label can only rely on user data and historical experience, and there is no unified and scientific prediction method. Moreover, the RFM model segmentation results include too many user groups, making it difficult to accurately position each user group, which makes it difficult for some companies with limited funds and resources to formulate effective marketing strategies for each segmented user group and to efficiently iterate products or marketing methods. As the number of users increases and user data becomes more diverse, the accuracy and objectivity of the potential labels of users predicted by traditional methods are getting worse and worse.

[0050] In order to solve the above technical problems, the present invention proposes a potential label mining method. The following takes a potential label mining device as an example to illustrate the implementation process of the potential label mining method provided by the embodiment of the present invention. It can be understood that the execution subject of the potential label mining method provided by the embodiment of the present invention includes but is not limited to the potential label mining device.

[0051] Figure 2 A flowchart of a potential tag mining method provided by an embodiment of the present invention is shown in FIG. Figure 2 As shown, the method includes:

[0052] Step 101: pre-process the original data of the user to be mined to generate user location data.

[0053] Step 102: Generate high-frequency location features based on user location data.

[0054] Step 103: Match the high-frequency location features with the acquired target location features of the target user through a tag mining model to generate potential tags for the user to be mined.

[0055] In the technical solution provided by the embodiment of the present invention, the original data of the user to be mined is preprocessed to generate user location data; high-frequency location features are generated based on the user location data; the high-frequency location features and the target location features of the target user obtained are matched through a label mining model to generate potential labels for the user to be mined, which can accurately locate the user type and assign accurate weights to each indicator, thereby improving the accuracy of the label mining model and formulating effective marketing strategies.

[0056] Figure 3 A flowchart of another potential tag mining method provided by an embodiment of the present invention is as follows: Figure 3 As shown, the method includes:

[0057] Step 201: divide the original data according to the user identification code of the user to be mined, and generate the original data corresponding to the user identification code.

[0058] In the embodiment of the present invention, each step is performed by a potential label mining device.

[0059] In the embodiment of the present invention, taking the user identification code as the user epidemic prevention code as an example, the user epidemic prevention code can uniquely represent a user, and each time the user epidemic prevention code of the user to be mined is displayed, the background server obtains raw data including the current location, and the raw data is not limited to date, time and location coordinates, wherein the location coordinates include longitude and latitude. Since raw data is collected each time each user epidemic prevention code is displayed, a large amount of raw data mixed with multiple users will be collected, and the raw data will be divided according to the user epidemic prevention code to generate multiple raw data corresponding to the user epidemic prevention code of each user.

[0060] It is understandable that other forms of user identification codes may be used to uniquely identify a user, and the original data may be obtained by using the user identification code. The embodiment of the present invention does not limit the specific form of the user identification code.

[0061] Step 202: clean the original data corresponding to the user identification code to generate user location data.

[0062] In the embodiment of the present invention, the user location data includes time and location coordinates.

[0063] In the embodiment of the present invention, cleaning the raw data includes filtering outliers and missing values ​​in the raw data and performing feature encoding and regularization on the filtered raw data. Specifically, the raw data in which the number of the same position coordinates in a specified time period is less than 5 is filtered. The frequency of the position coordinates is too low and has no reference value, and is an outlier; the raw data corresponding to the null value in the raw data is filtered. The raw data is a missing value, and if used for subsequent calculations, it will have a negative impact on the accuracy of the mining label.

[0064] As an optional solution, the filtered raw data is encoded by a one-hot encoding algorithm to generate a raw data encoding vector; the raw data encoding vector is regularized to generate user location data.

[0065] It is understandable that other methods may be used to clean the original data, and the present invention does not limit the specific implementation method of cleaning the original data.

[0066] In the embodiment of the present invention, the cleaned raw data, i.e., user location data, can better describe the user behavior characteristics, and avoid the negative impact of location information noise generated by extremely low-frequency location coordinates on subsequent calculation results, thereby improving the accuracy of the label mining model.

[0067] Step 203: Generate high-frequency location features based on user location data.

[0068] In the embodiment of the present invention, the high-frequency position feature includes a high-frequency position to be mined.

[0069] Step 203 specifically includes:

[0070] Step 2031: Statistics are performed on the time and location coordinates according to a preset period to generate the location frequency of the user to be mined at each location coordinate.

[0071] In the embodiment of the present invention, the preset period can be set according to actual conditions, and the embodiment of the present invention does not limit this. As an optional solution, the preset period is 30 days between May 1 and May 30.

[0072] For example: the preset period is 30 days between May 1st and May 30th. The number of users to be mined at location coordinate A from May 1st to May 30th is counted as 50, the number at location coordinate B is 30, and the number at location coordinate C is 20. Then the location frequency of the user to be mined at location coordinate A is 0.5, the location frequency at location coordinate B is 0.3, and the location frequency at location coordinate C is 0.2.

[0073] Step 2032: Sort the position coordinates according to position frequency.

[0074] In the embodiment of the present invention, the sorting method is from large to small or from small to large, and the embodiment of the present invention does not limit the sorting method.

[0075] For example: the location frequency of the user to be mined at location coordinate A is 0.5, the location frequency at location coordinate B is 0.3, and the location frequency at location coordinate C is 0.2. The location coordinates are sorted from large to small according to the location frequency. The sorted location coordinates are location coordinate A, location coordinate B, and location coordinate C.

[0076] Step 2033: Select the high-frequency position to be mined of the first specified sorting position.

[0077] In the embodiment of the present invention, the first designated sorting position can be set according to actual conditions, for example, if the position coordinates are sorted from large to small according to the position frequency, the first sorting position can be set to take the first ten position coordinates as the high-frequency positions to be mined. The embodiment of the present invention is not limited to this.

[0078] In the embodiment of the present invention, the high-frequency location features also include high-frequency administrative codes to be mined.

[0079] Step 203 specifically also includes:

[0080] Step 3031, query the administrative code corresponding to the location coordinates.

[0081] For example: the administrative code corresponding to location coordinate A is 300000, the administrative code corresponding to location coordinate A' is 300000, the administrative code corresponding to location coordinate B is 300001, and the administrative code corresponding to location coordinate B' is 300001.

[0082] Step 3032: Count the number of each administrative code and sort the administrative codes according to the counted number.

[0083] In the embodiment of the present invention, the same administrative codes are merged and the number of each administrative code is counted. For example, the number of administrative codes 300000 is 50, and the number of administrative codes 300001 is 20.

[0084] In the embodiment of the present invention, the sorting method is from large to small or from small to large, and the embodiment of the present invention does not limit the sorting method.

[0085] For example, the number of administrative codes 300000 is 50, and the number of administrative codes 300001 is 20. Sort the administrative codes from large to small, and the sorted administrative codes are administrative code 300000 and administrative code 300001.

[0086] Step 3033: Select the high-frequency administrative code to be mined in the second specified sorting position.

[0087] In the embodiment of the present invention, the second designated sorting position can be set according to the actual situation. For example, if the administrative codes are sorted from large to small, the second sorting position can be set to take the first ten administrative code labels as the high-frequency administrative codes to be mined. The embodiment of the present invention is not limited to this.

[0088] In the embodiment of the present invention, the high-frequency position features also include difference features to be mined.

[0089] Step 4031: Select the highest frequency position to be mined from the sorted position coordinates.

[0090] In the embodiment of the present invention, the position coordinates with the highest position frequency are selected from the sorted position coordinates as the highest frequency position to be mined.

[0091] Step 4032: Generate the difference feature to be mined according to the target highest frequency position and the highest frequency position to be mined by using the weighted Euclidean distance formula.

[0092] Specifically, through the Euclidean distance formula The target highest frequency position and the to-be-mined highest frequency position are calculated to generate the Euclidean distance between the target highest frequency position and the to-be-mined highest frequency position; the Euclidean distance is multiplied by the calculated difference weight to obtain the difference feature to be mined.

[0093] Among them, (x1, y1) is the position coordinate of the target highest frequency position, (x2, y2) is the position coordinate of the highest frequency position to be mined, and D is the Euclidean distance between the target highest frequency position and the highest frequency position to be mined.

[0094] In the embodiment of the present invention, the target highest frequency position is the highest frequency position of the target user with the user tag, the target position feature of the target user is obtained in advance, and the target position feature includes the target highest frequency position. The number of the highest frequency positions of the user to be mined is divided by the total number of position coordinates to obtain the difference weight.

[0095] In the embodiment of the present invention, the high-frequency location features include precise latitude and longitude location coordinates and high-frequency administrative codes, both of which have the main characteristics of being dynamic, massive, and time-space related. The subsequent tag mining model is used to predict the potential tags of the users to be mined, and the available social data is used for in-depth analysis, creating a visible value channel for business operations and creating a model for data application in other areas of government affairs, with a wide range of market prospects for output capabilities in provinces across the country.

[0096] Step 204: Match the high-frequency location features with the acquired target location features of the target user through a tag mining model to generate potential tags for the user to be mined.

[0097] Specifically, through the Gaussian Bayes classifier Match the high-frequency position features with the target position features to generate potential tags for the users to be mined. i |y) is the matching probability of the user to be mined based on the high-frequency location features under the target location features of the target user, x i is the high-frequency position feature, y is the target position feature and the target position feature corresponds to the user label of the target user, μ y is the mean of the high-frequency position features, σ y is the standard deviation of the high-frequency position feature. y and σ y is pre-computed based on the maximum likelihood estimation method.

[0098] In the embodiment of the present invention, by comparing the matching probabilities of the users to be mined, the user tag of the target user corresponding to the maximum matching probability is determined as the potential tag of the user to be mined.

[0099] In the embodiment of the present invention, a Gaussian Bayesian classifier is used to calculate the matching probability, which has a small error rate and a fast calculation speed, can handle multi-classification tasks, and is suitable for incremental training. The label mining model matches and classifies the users to be mined and predicts potential labels, so that the potential labels are highly correlated with behavioral changes and fluctuations in business indicators, revealing deeper user cognition from the potential labels.

[0100] In the embodiment of the present invention, if the number of high-frequency position features is too large, in order to avoid the problem of overfitting caused by excessively high model dimensions, feature selection may be performed on the high-frequency position features.

[0101] As an optional solution, feature selection specifically includes:

[0102] Step 301: Evaluate the high-frequency position features by specifying a classifier, and generate feature weights corresponding to each dimension of the high-frequency position features.

[0103] Specifically, the high-frequency position features are input into a specified classifier for evaluation, and the feature weights corresponding to each dimension of the high-frequency position features are output.

[0104] As an optional solution, the classifier is specified as a random forest algorithm. It is understandable that other classifiers may also be used to evaluate the high-frequency position features, which is not limited in the embodiment of the present invention.

[0105] In the embodiment of the present invention, the larger the feature weight corresponding to the high-frequency position feature is, the greater the relevance of the dimensional feature for model construction is, and the higher its importance is.

[0106] Step 302: sort the feature weights, and select the high-frequency position features corresponding to the feature weights of the third specified sorting positions.

[0107] In the embodiment of the present invention, the sorting method is from large to small or from small to large, and the embodiment of the present invention does not limit the sorting method.

[0108] In the embodiment of the present invention, the third designated sorting position can be set according to actual conditions, and high-frequency position features with higher importance are selected, which is not limited in the embodiment of the present invention.

[0109] Furthermore, if the dimension of the high-frequency position feature after feature processing is still high, in order to avoid the problem of overfitting caused by the model dimension being too high, the high-frequency position feature can be processed by dimensionality reduction. Specifically, by specifying a dimensionality reduction method, multiple high-frequency position features are converted into a low-dimensional space to generate key indicator features after dimensionality reduction. The dimensionality reduction method includes changing the original feature space through the relationship between feature attributes (for example: combining different attributes to obtain new attributes), extracting the main features, and thus achieving feature dimensionality reduction. Among them, the specified dimensionality reduction method is the principal component analysis (PCA) algorithm, and other dimensionality reduction methods can also be used to reduce the dimensionality of high-frequency position features, which is not limited in the embodiments of the present invention.

[0110] In the embodiment of the present invention, collecting high-frequency position features can improve the performance of the label mining model, increase prediction accuracy, reduce the time cost of later label mining model maintenance, and improve the reusability and response speed of the label mining model. The high-dimensional high-frequency position features are transformed into low-frequency position features using linear discriminant analysis, and then the low-dimensional high-frequency position features are input into the label mining model for clustering, which ensures high cohesion and low coupling of the class, thereby solving the problem that there are too many user segment groups and cannot provide leading guidance for operations. In addition, the present invention defines feature importance through feature selection, focusing on solving the problems of non-standard score segment and score judgment and inaccurate indicator weight calculation brought about by traditional customer segmentation methods.

[0111] In the embodiment of the present invention, before matching the high-frequency location features with the acquired target location features of the target user through the tag mining model and generating the potential tags of the user to be mined, it is necessary to pre-build the tag mining model. The specific construction process of the tag mining model is described below:

[0112] Step 401: Obtain target location features of the target user, where the target location features include target high-frequency location, target high-frequency administrative code, target difference features, and training labels.

[0113] In an embodiment of the present invention, the target location features of the target user are acquired in advance, wherein the target high-frequency location, target high-frequency administrative code and target difference features are acquired in the same manner as the high-frequency location to be mined, the high-frequency administrative code to be mined and the difference features to be mined, and are not described in detail here; the training label is a user label pre-set for the target user, and the training label is used to input a supervised learning method for training to construct a label mining model.

[0114] Step 402: Through supervised learning methods, target high-frequency positions, target high-frequency administrative codes, target difference features and training labels are trained to build a label mining model.

[0115] In the embodiment of the present invention, the target high-frequency position, target high-frequency administrative code, target difference feature and training label are input into the supervised training method for training to build a label mining model. As an optional solution, the supervised learning method is a Gaussian Bayes classifier, which has a small error rate and fast calculation speed, can handle multi-classification tasks, and is suitable for incremental training.

[0116] Furthermore, after generating potential tags for the users to be mined, operation strategies can be specified for the users based on the potential tags. For example, a reading application uses off-site promotion by users to attract new users by inviting friends to get red envelopes. It adds tags indicating that users have attracted new users, and adds tags indicating that users have not shared new users. The product operation department continuously incentivizes and stimulates users with the tag "no new user behavior", such as pop-up stimulation, increasing the amount of rewards, etc., so as to promote users with no new user behavior to spread the word, continuously incentivize this group of users, and achieve the goal of "every user is a promoter", thereby increasing the user acquisition rate. In addition, all tag strategies can be regularly back-traced to understand the conversion, effect, and impact of each operation strategy for tagged users, so as to continuously optimize the strategy and complete the iteration of user operation strategies driven by user tags.

[0117] In the embodiment of the present invention, the user epidemic prevention code of the user to be mined is taken as the core, the geographic location data of the user to be mined is mined, the behavior path of the user to be mined is quantified, the high-frequency location of the user to be mined is counted from two dimensions of location coordinates and administrative code, and matched with the high-frequency location of the target user, and the displacement behavior habits of the user to be mined are analyzed, and the label prediction of the user to be mined is realized, thereby helping enterprises to obtain user characteristics more concretely through different dimensions and achieve breakthroughs in intelligent operation and precision marketing from the user end.

[0118] In the technical solution of the potential tag mining method provided by the embodiment of the present invention, the original data of the user to be mined is preprocessed to generate user location data; high-frequency location features are generated based on the user location data; the high-frequency location features and the target location features of the target user obtained are matched through the tag mining model to generate potential tags for the user to be mined, which can accurately locate the user type and assign accurate weights to each indicator, thereby improving the accuracy of the tag mining model and formulating effective marketing strategies.

[0119] Figure 4 A schematic diagram of a potential tag mining device provided by an embodiment of the present invention is provided, wherein the device is used to execute the potential tag mining method described above. Figure 4 As shown, the device includes: a preprocessing unit 11, a generating unit 12 and a matching unit 13.

[0120] The preprocessing unit 11 is used to preprocess the acquired original data of the user to be mined to generate user location data.

[0121] The generating unit 12 is used to generate high-frequency position features according to the user position data.

[0122] The matching unit 13 is used to match the high-frequency position features with the acquired target position features of the target user through a tag mining model, and generate potential tags of the user to be mined.

[0123] In the embodiment of the present invention, the preprocessing unit 11 is specifically used to divide the original data according to the user identification code of the user to be mined, generate the original data corresponding to the user identification code; clean the original data corresponding to the user identification code, and generate the user location data.

[0124] In an embodiment of the present invention, the generation unit 12 is specifically used to count the time and location coordinates according to a preset period to generate the location frequency of the user to be mined at each location coordinate; sort the location coordinates according to the location frequency; and select the high-frequency position to be mined at the first specified sorting position.

[0125] In an embodiment of the present invention, the generation unit 12 is specifically used to query the administrative code corresponding to the location coordinates; count the number of each administrative code, and sort the administrative codes according to the counted number; and select the high-frequency administrative code to be mined at the second specified sorting position.

[0126] In the embodiment of the present invention, the generating unit 12 is further used to select the highest frequency position to be mined from the sorted position coordinates; and generate the difference feature to be mined according to the target highest frequency position and the highest frequency position to be mined by using the weighted Euclidean distance formula.

[0127] In the embodiment of the present invention, the device further includes: an evaluation unit 14 and a selection unit 15 .

[0128] The evaluation unit 14 is used to evaluate the high-frequency position features through a specified classifier, and generate feature weights corresponding to each dimension of the high-frequency position features.

[0129] The selection unit 15 is used to sort the feature weights and select the high-frequency position feature corresponding to the feature weight of the third specified sorting position.

[0130] In the embodiment of the present invention, the device further includes a conversion unit 16 .

[0131] The conversion unit 16 is used to perform low-dimensional space conversion on multiple high-frequency position features by specifying a dimensionality reduction method to generate key indicator features after dimensionality reduction.

[0132] In the embodiment of the present invention, the device further includes: an acquisition unit 17 and a training unit 18.

[0133] The acquisition unit 17 is used to acquire the target location characteristics of the target user, where the target location characteristics include a target high-frequency location, a target high-frequency administrative code, a target difference characteristic and a training label.

[0134] The training unit 18 is used to train the target high-frequency position, target high-frequency administrative code, target difference feature and training label through supervised learning method to build a label mining model.

[0135] In the scheme of the embodiment of the present invention, the original data of the user to be mined is preprocessed to generate user location data; high-frequency location features are generated based on the user location data; the high-frequency location features and the target location features of the target user obtained are matched through a label mining model to generate potential labels for the user to be mined, which can accurately locate the user type and assign accurate weights to each indicator, thereby improving the accuracy of the label mining model and formulating effective marketing strategies.

[0136] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer device, and specifically, the computer device may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0137] An embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory is used to store information including program instructions, and the processor is used to control the execution of the program instructions. When the program instructions are loaded and executed by the processor, the steps of the embodiment of the above-mentioned potential label mining method are implemented. For a specific description, please refer to the embodiment of the above-mentioned potential label mining method.

[0138] Reference below Figure 5 , which shows a schematic diagram of the structure of a computer device 600 suitable for implementing an embodiment of the present application.

[0139] like Figure 5 As shown, the computer device 600 includes a central processing unit (CPU) 601, which can perform various appropriate operations and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 to a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the computer device 600 are also stored. The CPU 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0140] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal feedback device (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 606 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed as needed as the storage section 608.

[0141] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program tangibly contained on a machine-readable medium, and the computer program includes program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 609, and / or installed from the removable medium 611.

[0142] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0143] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0144] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0145] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0146] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0147] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0148] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0149] The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0150] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0151] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A potential tag mining method, characterized in that: The method comprises: Preprocess the original data of the users to be mined to generate user location data; Generate high-frequency location features according to the user location data, wherein the high-frequency location features include high-frequency locations to be mined and high-frequency administrative codes to be mined; Matching the high-frequency location features with the acquired target location features of the target user through a tag mining model to generate potential tags for the user to be mined; The step of matching the high-frequency location features with the acquired target location features of the target user through a tag mining model to generate potential tags of the user to be mined includes: The matching probabilities of the users to be mined are compared by using a Gaussian Bayes classifier, and the user tag of the target user corresponding to the maximum matching probability is determined as the potential tag of the user to be mined, wherein the matching probability is the matching probability of the user to be mined based on the high-frequency location feature under the target location feature condition of the target user; The user location data includes time and location coordinates; The step of generating high-frequency location features according to the user location data includes: The time and location coordinates are counted according to a preset period to generate the location frequency of the user to be mined at each location coordinate; sorting the position coordinates according to the position frequency; Selecting the high-frequency position to be mined of the first specified sorting position; Query the administrative code corresponding to the location coordinates; Counting the number of each administrative code, and sorting the administrative codes according to the counted number; Select the high-frequency administrative code to be mined in the second designated sorting position; Before matching the high-frequency location features with the acquired target location features of the target user through the tag mining model to generate the potential tags of the user to be mined, the method further includes: Obtaining target location features of the target user, wherein the target location features include a target high-frequency location, a target high-frequency administrative code, a target difference feature, and a training label; The label mining model is constructed by training the target high-frequency positions, target high-frequency administrative codes, target difference features and training labels through a supervised learning method.

2. The potential tag mining method according to claim 1, characterized in that: The preprocessing of the original data of the user to be mined to generate user location data includes: According to the user identification code of the user to be mined, the original data is divided to generate original data corresponding to the user identification code; The original data corresponding to the user identification code is cleaned to generate user location data.

3. The potential tag mining method according to claim 1, characterized in that: The high-frequency position features also include difference features to be mined; The target position feature includes the target highest frequency position; Generating high-frequency location features according to the user location data includes: Select the highest frequency position to be mined from the sorted position coordinates; The difference feature to be mined is generated according to the target highest frequency position and the highest frequency position to be mined by using a weighted Euclidean distance formula.

4. The potential tag mining method according to claim 1, characterized in that: Before matching the high-frequency location features with the acquired target location features of the target user through the tag mining model to generate the potential tags of the user to be mined, the method further includes: By specifying a classifier, the high-frequency position feature is evaluated to generate a feature weight corresponding to each dimension of the high-frequency position feature; The feature weights are sorted, and high-frequency position features corresponding to the feature weights of the third specified sorting positions are selected.

5. The potential tag mining method according to claim 4, characterized in that: After the feature weights are sorted and the high-frequency position features corresponding to the feature weights of the third specified sorting positions are selected, the method further includes: By specifying a dimensionality reduction method, a plurality of high-frequency position features are transformed into a low-dimensional space to generate key indicator features after dimensionality reduction.

6. A potential tag mining device, characterized in that: The device comprises: A preprocessing unit, used to preprocess the original data of the user to be mined to generate user location data; A generating unit, configured to generate high-frequency location features according to the user location data, wherein the high-frequency location features include high-frequency locations to be mined and high-frequency administrative codes to be mined; A matching unit, used for matching the high-frequency location features with the acquired target location features of the target user through a tag mining model to generate a potential tag of the user to be mined; The matching unit is specifically used to compare the matching probabilities of the users to be mined through a Gaussian Bayes classifier, and determine the user tag of the target user corresponding to the maximum matching probability as the potential tag of the user to be mined, wherein the matching probability is the matching probability of the user to be mined based on the high-frequency position feature under the target position feature condition of the target user; The user location data includes time and location coordinates; The generating unit is specifically used to count the time and location coordinates according to a preset period, generate the location frequency of the user to be mined at each location coordinate; sort the location coordinates according to the location frequency; select the high-frequency location to be mined at the first specified sorting position; query the administrative code corresponding to the location coordinate; count the number of each administrative code, and sort the administrative codes according to the counted number; select the high-frequency administrative code to be mined at the second specified sorting position; The device also includes: an acquisition unit and a training unit; The acquisition unit is used to acquire the target location characteristics of the target user, the target location characteristics including the target high-frequency location, the target high-frequency administrative code, the target difference characteristics and the training label; The training unit is used to train the target high-frequency position, target high-frequency administrative code, target difference characteristics and training labels through a supervised learning method to build a label mining model.

7. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the potential label mining method as described in any one of claims 1 to 5 is implemented.

8. A computer device comprising a memory and a processor, wherein the memory is used to store information including program instructions, and the processor is used to control the execution of the program instructions, characterized in that: When the program instructions are loaded and executed by the processor, the potential label mining method described in any one of claims 1 to 5 is implemented.

Images