Method and device for generating user tags

Abstract

The present invention discloses a method and device for generating user labels. A specific implementation of the method includes: dividing all users into multiple user groups; for each target classification feature, training a binary classification model with the first target user group as a positive sample and the remaining users as negative samples, determining the first probability that the label of each user in the second target user group set is the target classification feature; training a label model based on the target classification feature of each user in the second target user group set and the first probability that the label is each target classification feature; determining the second probability that the label of the user to be identified is each target classification feature based on each of the binary classification models, and generating a label for the user to be identified based on the second probability and the label model. This implementation can learn the behavioral characteristics of all users during the model training process, so that the trained label model can be applied to the entire user group, thereby improving the accuracy of user label generation.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and device for generating user tags. Background Art

[0002] In the existing solutions to label problems, the main approach is to transform the problem into a simple classification problem. For example, when classifying and analyzing users, users who only meet the characteristics of each target classification are usually screened out to define positive and negative samples for model training (e.g., users who have only purchased self-operated products are 1, and users who have only purchased POP products are 0), and then a machine learning classification algorithm is selected for modeling.

[0003] In the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art:

[0004] When defining positive and negative samples, most of the samples will be lost, such as user samples that meet multiple target classification features at the same time or do not meet all target classification features. It is impossible to learn the characteristic information of all users, resulting in insufficient model training. When the model is applied to the entire user group for analysis, the model effect is greatly reduced. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a method and device for generating user tags, which can learn the behavioral characteristics of all users during the model training process, so that the trained tag model can be applied to the entire user group, thereby improving the accuracy of user tag generation.

[0006] To achieve the above object, according to one aspect of an embodiment of the present invention, a method for generating a user tag is provided, comprising:

[0007] According to the user classification characteristics, all users are divided into multiple user groups; among all users, users who meet the target classification characteristics are the first target user group; users who only meet the target classification characteristics are the second target user group;

[0008] For each of the target classification features, a binary classification model is trained with the corresponding first target user group as a positive sample and the remaining users among all users as negative samples; a first probability that the label of each user in the second target user group set is the target classification feature is determined according to the behavioral characteristics of each user in the second target user group set and the binary classification model;

[0009] Training a label model according to the target classification feature of each user in the second target user group set and the first probability that the label is each target classification feature;

[0010] The second probability that the label of the user to be identified is each target classification feature is determined according to each of the two classification models, and the label of the user to be identified is generated according to the second probability and the label model.

[0011] Optionally, generating a tag of the user to be identified according to the second probability and the tag model includes:

[0012] The second probability that the label of the user to be identified is each target classification feature is input into the label model to obtain the third probability that the label of the user to be identified is each target classification feature; according to the third probability that the label of the user to be identified is each target classification feature, the target classification feature or target classification feature list of the user to be identified is determined.

[0013] Optionally, the training features of the binary classification model are determined based on the target classification features.

[0014] Optionally, the classification feature is a shopping channel; the user's behavior feature includes at least one of the following: browsing items, adding items to a shopping cart, collecting items, and following items.

[0015] According to a second aspect of an embodiment of the present invention, there is provided a device for generating a user tag, including:

[0016] A user group division module, which divides all users into multiple user groups according to user classification characteristics; among all users, users who meet the target classification characteristics are the first target user group; users who only meet the target classification characteristics are the second target user group;

[0017] A binary classification model training module, for each of the target classification features, using the corresponding first target user group as a positive sample and the remaining users among all users as negative samples to train a binary classification model; determining a first probability that the label of each user in the second target user group set is the target classification feature according to the behavioral characteristics of each user in the second target user group set and the binary classification model;

[0018] A label model training module, which trains a label model according to the target classification feature of each user in the second target user group set and the first probability that the label is each target classification feature;

[0019] The label generation module determines the second probability that the label of the user to be identified is each target classification feature according to each of the two classification models, and generates the label of the user to be identified according to the second probability and the label model.

[0020] Optionally, the tag generation module generates a tag of the user to be identified according to the second probability and the tag model, including:

[0021] The second probability that the label of the user to be identified is each target classification feature is input into the label model to obtain the third probability that the label of the user to be identified is each target classification feature; according to the third probability that the label of the user to be identified is each target classification feature, the target classification feature or target classification feature list of the user to be identified is determined.

[0022] Optionally, the training features of the binary classification model are determined based on the target classification features.

[0023] Optionally, the classification feature is a shopping channel; the user's behavior feature includes at least one of the following: browsing items, adding items to a shopping cart, collecting items, and following items.

[0024] According to a third aspect of an embodiment of the present invention, there is provided an electronic device for generating a user tag, including:

[0025] one or more processors;

[0026] a storage device for storing one or more programs,

[0027] When the one or more programs are executed by the one or more processors, the one or more processors implement the method provided in the first aspect of the embodiment of the present invention.

[0028] According to a fourth aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method provided by the first aspect of the embodiment of the present invention is implemented.

[0029] An embodiment of the above invention has the following advantages or beneficial effects: for each target classification feature, a binary classification model is trained by using the first target user group that meets the target classification feature as positive samples and the remaining users as negative samples, and a label model is trained based on the binary classification model. The behavioral characteristics of all users can be learned during the model training process, so that the trained label model can be applied to the entire user group, thereby improving the accuracy of user label generation.

[0030] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0032] Figure 1 is a schematic diagram of the main process of the method for generating user tags according to an embodiment of the present invention;

[0033] Figure 2 is a schematic diagram of a method for generating user tags in some embodiments of the present invention;

[0034] Figure 3 is a schematic diagram of a method for generating user tags in other embodiments of the present invention;

[0035] Figure 4 is a schematic diagram of main modules of a device for generating user tags according to an embodiment of the present invention;

[0036] Figure 5 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0037] Figure 6 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0039] According to one aspect of an embodiment of the present invention, a method for generating a user tag is provided.

[0040] Figure 1 FIG. 1 is a schematic diagram of the main process of the method for generating user tags according to an embodiment of the present invention. Figure 1 As shown, the method for generating user tags includes: step S101, step S102, step S103 and step S104.

[0041] Step S101, according to user classification characteristics, all users are divided into multiple user groups; among all users, users who meet the target classification characteristics are the first target user group; users who only meet the target classification characteristics are the second target user group.

[0042] Classification features can be selectively set according to actual conditions. Target classification features refer to a specific classification feature content.

[0043] Taking the e-commerce field as an example, all users can be divided into multiple user groups based on the user shopping channel as a classification feature. Assume that the user shopping channels include: platform self-operated and third-party cooperative stores. For the first target classification feature "platform self-operated", the first target user group is users who have a record of purchasing platform self-operated products, and the second target user group is users who only shop through the platform self-operated method. For the second target classification feature "third-party cooperative store", the first target user group is users who have a record of purchasing third-party cooperative store products, and the second target user group is users who only shop through third-party cooperative stores. Of course, the purchase behavior mentioned here can also be replaced by browsing behavior, add to cart behavior, attention behavior, collection behavior, click on product behavior, etc.

[0044] Taking the banking sector as an example, all users can be divided into multiple user groups based on the user's investment method as a classification feature. Assume that a bank's investment methods include: purchasing financial products, purchasing bonds, and purchasing funds. For the first target classification feature "purchasing financial products", the first target user group is users who have invested through "purchasing financial products", and the second target user group is users who only invest through "purchasing financial products". For the second target classification feature "purchasing bonds", the first target user group is users who have invested through "purchasing bonds", and the second target user group is users who only invest through "purchasing bonds". For the third target classification feature "purchasing funds", the first target user group is users who have invested through "purchasing funds", and the second target user group is users who only invest through "purchasing funds".

[0045] Step S102: for each of the target classification features, a binary classification model is trained with the corresponding first target user group as a positive sample and the remaining users among all users as negative samples; and a first probability that the label of each user in the second target user group set is the target classification feature is determined according to the behavioral characteristics of each user in the second target user group set and the binary classification model.

[0046] The user's behavioral characteristics refer to the characteristics that can reflect the user's label. Taking the e-commerce field as an example, the user's behavioral characteristics may include at least one of the following: browsing items, adding items to a purchase, collecting items, and following items. Taking the banking field as an example, the user's behavioral characteristics may include the purchase, following, or redemption of financial products, bonds, funds, etc. Taking public transportation as an example, the user's behavioral characteristics may include at least one of the following: the user's time to go to work or get off work, the user's choice of travel mode, the user's choice of travel fee settlement method, etc.

[0047] The actual label of each user in the positive sample is the target classification feature, and the actual label of each user in the negative sample is not the target classification feature. For each target classification feature, when training the binary classification model, the behavior characteristics of each user in the first target user group are used as the model input and the actual label of each user is used as the output for model training.

[0048] The second target user group set refers to the set of second target user groups corresponding to each target classification feature.

[0049] Take the aforementioned e-commerce scenario as an example. Assume that the user's shopping channels include: platform-owned and third-party cooperative stores. For the first target classification feature "platform-owned", the first target user group is users who have purchased platform-owned products, and the second target user group is users who only shop through the platform's own methods. For the second target classification feature "third-party cooperative stores", the first target user group is users who have purchased third-party cooperative stores, and the second target user group is users who only shop through third-party cooperative stores. The two second target user groups constitute the second target user group set.

[0050] In this step, for the first target classification feature "platform self-operated", a binary classification model is trained with users who have purchased platform self-operated products as positive samples and users who have not purchased platform self-operated products as negative samples, and the first probability of the label of each user in the second target user group set being purchased through platform self-operated is determined based on the binary classification model. For the second target classification feature "third-party cooperative store", a binary classification model is trained with users who have purchased third-party cooperative store products as positive samples and users who have not purchased third-party cooperative store products as negative samples, and the first probability of the label of each user in the second target user group set being purchased through third-party cooperative stores is determined based on the binary classification model.

[0051] Take the aforementioned banking scenario as an example. Assume that a bank's investment methods include: purchasing wealth management products, purchasing bonds, and purchasing funds. For the first target classification feature "purchasing wealth management products", the first target user group is users who have invested through "purchasing wealth management products", and the second target user group is users who only invest through "purchasing wealth management products". For the second target classification feature "purchasing bonds", the first target user group is users who have invested through "purchasing bonds", and the second target user group is users who only invest through "purchasing bonds". For the third target classification feature "purchasing funds", the first target user group is users who have invested through "purchasing funds", and the second target user group is users who only invest through "purchasing funds". The three second target user groups constitute the second target user group set.

[0052] In this step, for the first target classification feature "purchase of financial products", a binary classification model is trained with users who have made investments through "purchase of financial products" as positive samples and users who have not made investments through "purchase of financial products" as negative samples, and the first probability of the label of each user in the second target user group set being investment through "purchase of financial products" is determined based on the binary classification model. For the second target classification feature "purchase of bond products", a binary classification model is trained with users who have made investments through "purchase of bond products" as positive samples and users who have not made investments through "purchase of bond products" as negative samples, and the first probability of the label of each user in the second target user group set being investment through "purchase of bond products" is determined based on the binary classification model. For the third target classification feature "purchase of fund products", a binary classification model is trained with users who have made investments through "purchase of fund products" as positive samples and users who have not made investments through "purchase of fund products" as negative samples, and the first probability of the label of each user in the second target user group being investment through "purchase of fund products" is determined based on the binary classification model.

[0053] This step uses the behavioral characteristics of all users as training samples, so that the behavioral characteristics of all users can be learned during the model training process, so that the label model trained on this basis can be applied to the entire user group, thereby improving the accuracy of user label generation.

[0054] Step S103: training a label model according to the target classification feature of each user in the second target user group set and the first probability that the label is each target classification feature.

[0055] Exemplarily, assuming that the number of target classification features is two, the second target user group set includes: a second target user group 1 corresponding to the target classification feature 1, and a second target user group 2 corresponding to the target classification feature 2, the target classification feature 1 corresponds to the binary classification model 1, and the target classification feature 2 corresponds to the binary classification model 2. For each user in the second target user group 1, its target classification feature is the target classification feature 1, and its corresponding first probability includes the first probability determined by the binary classification model 1 and the first probability determined by the binary classification model 2. For each user in the second target user group 2, its target classification feature is the target classification feature 2, and its corresponding first probability includes the first probability determined by the binary classification model 1 and the first probability determined by the binary classification model 2.

[0056] In the embodiment of the present invention, the training features used in the binary classification model training and label model training processes are completely different. The training features in the binary classification model training process are whether the user meets the target classification features, and the training features in the label model training process are the first probabilities output by each binary classification model. In addition, the division of training samples in the embodiment of the present invention is based on model logic rather than random sampling, and users who meet a certain target classification feature are used as positive samples to train the binary classification model corresponding to the target classification feature. The present invention learns the feature information of the entire user group during the model training process, so that the trained label model can be applied to the entire user group to generate accurate user labels.

[0057] Step S104: determining the second probability that the label of the user to be identified is each target classification feature according to each of the binary classification models, and generating the label of the user to be identified according to the second probability and the label model.

[0058] In this step, the second probability that the label of the user to be identified is each target classification feature is determined according to each binary classification model, and then the second probability output by each binary classification model is input into the label model, and the label of the user to be identified is generated according to the output of the label model. The output of the label model can be a determined result, such as result information that identifies the label of the user to be identified as a certain target classification feature. Of course, the output of the label model can also be the probability that the label of the user to be identified is each target classification feature.

[0059] The number of the target classification features is two or more. Optionally, the number of the target classification features is two; training the label model according to the first probability of each user in the second target user group corresponding to each of the target classification features includes: training the label model with the first probability of each user in the second target user group corresponding to one target classification feature as a positive sample and with the first probability of each user in the second target user group corresponding to another target classification feature as a negative sample.

[0060] User behavior features are training features when training a binary classification model. Optionally, the training features of the binary classification model are determined based on the target classification features. When training the binary classification models corresponding to each target classification feature, the training features may be different. For example, when training a binary classification model corresponding to the target classification feature "purchase of low-value goods (i.e., goods with relatively low prices)", select behavioral features related to the purchase of low-value goods, such as the features of browsing, adding to carts (adding to shopping carts), following, and collecting when users purchase low-value goods; when training a binary classification model corresponding to the target classification feature "purchase of high-value goods (i.e., goods with relatively high prices)", select behavioral features related to the purchase of high-value goods, such as the features of browsing, adding to carts (adding to shopping carts), following, and collecting when users purchase high-value goods.

[0061] Optionally, generating a label of the user to be identified based on the second probability and the label model includes: inputting the second probability that the label of the user to be identified is each target classification feature into the label model to obtain a third probability that the label of the user to be identified is each target classification feature; and determining the target classification feature or a target classification feature list of the user to be identified based on the third probability that the label of the user to be identified is each target classification feature. This embodiment can accurately determine the user label or the target classification feature list of the label, and the slogan can be used to provide marketing incentives to the user later.

[0062] Figure 2 A schematic diagram of a method for generating user tags in an optional embodiment of the present invention is shown. Figure 2 In the optional embodiment shown, the user classification feature is the shopping channel, and the number of target classification features is three, namely: purchase of product X, purchase of product Y, and purchase of product Z. The method for generating user tags mainly includes five stages:

[0063] 1. Customer segmentation:

[0064] According to the user classification characteristics, all users are divided into five user groups: customer group K1, customer group K2, customer group K3, customer group K4, customer group K5, customer group K6 and customer group K7. Among them, customer group K1 is users who have bought product X, customer group K2 is users who have bought product Y, customer group K3 is users who have bought product Z, customer group K4 is users who have not bought products X, Y and Z, customer group K5 is users who have only bought product X, customer group K6 is users who have only bought product Y, and customer group K7 is users who have only bought product Z. There may be duplicate users among customer groups K1, K2 and K3, that is, users who have bought at least two of products X, Y and Z.

[0065] 2. Independent training:

[0066] Take customer group K1 as the positive sample, customer groups K2, K3, and K4 as negative samples, and train a binary classification model, which is model X. In the process of training model X, customer group K1 is marked as 1, customer groups K2, K3, and K4 are marked as 0, and the behavioral characteristics of users purchasing product X are used as input variables, and the machine learning binary classification algorithm is used for model training.

[0067] Take customer group K2 as the positive sample, customer groups K1, K3, and K4 as the negative samples, and train a binary classification model, which is model Y. In the process of training model Y, customer group K2 is marked as 1, customer groups K1, K3, and K4 are marked as 0, and the behavioral characteristics of users purchasing product Y are used as input variables. The machine learning binary classification algorithm is also used for model training.

[0068] Take customer group K3 as the positive sample, customer groups K1, K2, and K4 as negative samples, and train a binary classification model, namely model Z. In the process of training model Z, customer group K3 is marked as 1, customer groups K1, K2, and K4 are marked as 0, and the behavioral characteristics of users purchasing product Z are used as input variables. The machine learning binary classification algorithm is also used for model training.

[0069] 3. Independent Analysis:

[0070] Model X is used to determine the first probability that the label of each user in customer groups K5, K6, and K7 is to purchase product X.

[0071] Model Y is used to determine the first probability that the label of each user in customer groups K5, K6, and K7 is to purchase product Y.

[0072] Model Z is used to determine the first probability that the label of each user in customer groups K5, K6, and K7 is to purchase product Z.

[0073] The output probability of the binary classification model is used to quantify whether the user will purchase the corresponding category of goods.

[0074] 4. Model Fusion

[0075] Customer groups K5, K6 and K7 are used as training samples, and the training features are the first probabilities output by model X, model Y and model Z.

[0076] 5. Result generation:

[0077] Use the label model to analyze all the users to be identified and generate labels for the users to be identified. In this example, the output of the label model includes the probability of the user purchasing products in categories X, Y, and Z: P X , P Y , P Z .P X It is used to quantify the degree to which users label category X among the three categories X / Y / Z. Y It is used to quantify the degree to which users label the Y category among the three categories of X / Y / Z. Z It is used to quantify the degree to which users label the Z category among the three categories X / Y / Z.

[0078] Since the multi-classification model generally uses the softmax function, there is P X +P Y +P Z =1, take max(P X ,P Y ,P Z ) is the classification of user tags. X , PY , P Z Sort the categories of user tags by the value of to obtain a tag list for each user.

[0079] Compared with only using customer groups K1, K2, and K3 as training samples, this embodiment uses all users for model training, that is, uses customer groups K1, K2, K3, and 4 for model training, which can avoid the problem that the generated label model is not applicable to all users due to incomplete samples.

[0080] In addition, the present invention uses different training features when training each binary classification model. User behavior features corresponding to the target classification features are used when training each binary classification model. Compared with using systematic training features or randomly using training features each time training the model, the present invention can significantly improve the model analysis effect.

[0081] Figure 3 A schematic diagram of a method for generating user tags in an optional embodiment of the present invention is shown. Figure 3 In the optional embodiment shown, the user classification feature is the shopping channel, and the number of target classification features is two, namely: purchasing self-operated goods and purchasing POP (third-party stores settled in the platform) goods. The method of generating user tags mainly includes five stages:

[0082] 1. Customer segmentation:

[0083] According to user classification characteristics, all users are divided into five user groups: customer group A, customer group B, customer group C, customer group D and customer group E. Among them, customer group A is users who have bought self-operated products, customer group C is users who have bought POP products, customer group B is users who have not bought self-operated and POP products, customer group D is users who have only bought self-operated products, and customer group E is users who have only bought POP products. Customer groups A and C may have duplicate users, that is, users who have bought both self-operated and POP products.

[0084] 2. Independent training:

[0085] With customer group A as the positive sample and customer groups B and C as the negative samples, a binary classification model is trained, which is the self-operated model. In the process of training the self-operated model, customer group A is marked as 1, customer groups B and customer groups C are marked as 0, and the user's self-operated characteristics (i.e., the behavioral characteristics of purchasing self-operated products) are used as input variables, and the machine learning binary classification algorithm is used for model training.

[0086] With customer group C as the positive sample and customer groups A and C as the negative samples, a binary classification model is trained, which is the POP model. In the process of training the POP model, customer group C is marked as 1, customer groups B and A are marked as 0, and the user's POP features (i.e., the behavioral features of purchasing POP products) are used as input variables, and the machine learning binary classification algorithm is also used for model training.

[0087] 3. Independent Analysis:

[0088] The self-operated model is used to determine the first probability that each user in customer group D is labeled as purchasing self-operated products, and the first probability that each user in customer group E is labeled as purchasing self-operated products.

[0089] The POP model is used to determine the first probability that each user in customer group D has a label that purchases POP products, and the first probability that each user in customer group E has a label that purchases POP products.

[0090] 4. Model Fusion

[0091] In the process of training the label model, customer group D is marked as 1, customer group E is marked as 0, and the two first probabilities output by the self-operated model and the POP model are used as input variables. The machine learning binary classification algorithm is used to integrate the above two models to obtain the label model.

[0092] 5. Result generation:

[0093] The label model is used to analyze all users to be identified and generate labels for the users to be identified. In this example, the output result of the label model may include the probability of the user purchasing self-operated products and the probability of purchasing POP products, or may only include a probability value. When the probability value is closer to 1, it indicates that the current user to be identified is more inclined to purchase self-operated products, and when the probability value is closer to 0, it indicates that the current user to be identified is more inclined to purchase POP products.

[0094] Compared with only using customer groups A and C as training samples, this embodiment uses all users for model training, that is, using customer groups A, B and C for model training, which can avoid the problem that the generated label model is not applicable to all users due to incomplete samples.

[0095] In addition, the present invention uses different training features when training each binary classification model. User behavior features corresponding to the target classification features are used when training each binary classification model. Compared with using systematic training features or randomly using training features each time training the model, the present invention can significantly improve the model analysis effect.

[0096] According to a second aspect of an embodiment of the present invention, a device for implementing the above method is provided.

[0097] Figure 4 FIG. 1 is a schematic diagram of the main modules of the apparatus for generating user tags according to an embodiment of the present invention. Figure 4 As shown, the device 400 for generating a user tag includes:

[0098] The user group division module 401 divides all users into multiple user groups according to user classification characteristics; among all users, users who meet the target classification characteristics are the first target user group; users who only meet the target classification characteristics are the second target user group;

[0099] The binary classification model training module 402 trains a binary classification model for each of the target classification features, using the corresponding first target user group as a positive sample and the remaining users among all users as negative samples; and determines a first probability that the label of each user in the second target user group set is the target classification feature according to the behavioral characteristics of each user in the second target user group set and the binary classification model;

[0100] The label model training module 403 trains the label model according to the target classification feature of each user in the second target user group set and the first probability that the label is each target classification feature;

[0101] The label generation module 404 determines the second probability that the label of the user to be identified is each target classification feature according to each of the two classification models, and generates the label of the user to be identified according to the second probability and the label model.

[0102] Optionally, the tag generation module generates a tag of the user to be identified according to the second probability and the tag model, including:

[0103] The second probability that the label of the user to be identified is each target classification feature is input into the label model to obtain the third probability that the label of the user to be identified is each target classification feature; according to the third probability that the label of the user to be identified is each target classification feature, the target classification feature or target classification feature list of the user to be identified is determined.

[0104] Optionally, the training features of the binary classification model are determined based on the target classification features.

[0105] Optionally, the classification feature is a shopping channel; the user's behavior feature includes at least one of the following: browsing items, adding items to a shopping cart, collecting items, and following items.

[0106] According to a third aspect of an embodiment of the present invention, there is provided an electronic device for generating a user tag, including:

[0107] one or more processors;

[0108] a storage device for storing one or more programs,

[0109] When the one or more programs are executed by the one or more processors, the one or more processors implement the method provided in the first aspect of the embodiment of the present invention.

[0110] According to a fourth aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method provided by the first aspect of the embodiment of the present invention is implemented.

[0111] Figure 5 An exemplary system architecture 500 is shown to which the method or apparatus for generating user tags according to the embodiments of the present invention may be applied.

[0112] like Figure 5 As shown, system architecture 500 may include terminal devices 501, 502, 503, a network 504 and a server 505. Network 504 is used to provide a medium for communication links between terminal devices 501, 502, 503 and server 505. Network 504 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0113] Users can use terminal devices 501, 502, 503 to interact with server 505 through network 504 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 501, 502, 503, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0114] The terminal devices 501 , 502 , and 503 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0115] The server 505 may be a server that provides various services, such as a backend management server (only an example) that provides support for shopping websites browsed by users using the terminal devices 501, 502, and 503. The backend management server may analyze and process the received data such as product information query requests, and feed back the processing results (such as target push information, product information - only an example) to the terminal device.

[0116] It should be noted that the method for generating a user tag provided in the embodiment of the present invention is generally executed by the server 505 , and accordingly, the device for generating a user tag is generally disposed in the server 505 .

[0117] It should be understood that Figure 5The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0118] Reference below Figure 6 , which shows a schematic diagram of the structure of a computer system 600 of a terminal device suitable for implementing an embodiment of the present invention. Figure 6 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0119] like Figure 6 As shown, the computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the system 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.

[0120] 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 display (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 605 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 into the storage section 608 as needed.

[0121] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the central processing unit (CPU) 601, the above-mentioned functions defined in the system of the present invention are executed.

[0122] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0123] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0124] The modules involved in the embodiments of the present invention can be implemented by software or hardware. The modules described can also be set in a processor, for example, can be described as: a processor includes: a user group division module, according to the user classification characteristics, all users are divided into multiple user groups; wherein, the users who meet the target classification characteristics among all users are the first target user group; the users who only meet the target classification characteristics are the second target user group; a binary classification model training module, for each of the target classification characteristics, the binary classification model is trained with the corresponding first target user group as a positive sample and the remaining users among all users as negative samples; according to the behavioral characteristics of each user in the second target user group set and the binary classification model, the first probability that the label of each user in the second target user group set is the target classification feature is determined; a label model training module, according to the target classification feature of each user in the second target user group set and the first probability that the label is each target classification feature, the label generation module, according to each of the binary classification models, determines the second probability that the label of the user to be identified is each target classification feature, and generates the label of the user to be identified according to the second probability and the label model. The names of these modules do not, in some cases, limit the modules themselves. For example, the user group segmentation module can also be described as a "module for generating labels for users to be identified based on the second probability and the label model."

[0125] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently and not be assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by one of the devices, the device includes: according to user classification features, all users are divided into multiple user groups; wherein, users among all users who meet the target classification features are the first target user group; users who only meet the target classification features are the second target user group; for each of the target classification features, a binary classification model is trained with the corresponding first target user group as a positive sample and the remaining users among all users as negative samples; according to the behavioral features of each user in the second target user group set and the binary classification model, the first probability that the label of each user in the second target user group set is the target classification feature is determined; according to the target classification features of each user in the second target user group set and the first probability that the label is each target classification feature, a label model is trained; according to each of the binary classification models, the second probability that the label of the user to be identified is each target classification feature is determined, and the label of the user to be identified is generated according to the second probability and the label model.

[0126] According to the technical solution of the embodiment of the present invention, the binary classification model is trained with the user group that meets the target classification characteristics as positive samples and the remaining users as negative samples. The probability of each user in the second target user group meeting the target classification characteristics is determined according to the binary classification model. The label model is trained according to the probability of each user in the second target user group corresponding to each target classification characteristic. The behavioral characteristics of all users can be learned during the model training process, so that the trained label model can be applied to the entire user group, thereby improving the accuracy of user label generation.

[0127] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for generating user tags, characterized in that: include: According to the user classification characteristics, all users are divided into multiple user groups; among all users, users who meet the target classification characteristics are the first target user group; users who only meet the target classification characteristics are the second target user group; For each of the target classification features, a binary classification model is trained with the corresponding first target user group as a positive sample and the remaining users among all users as negative samples; a first probability that the label of each user in the second target user group set is the target classification feature is determined according to the behavioral features of each user in the second target user group set and the binary classification model; when the number of the target classification features is two or more, different training features are used when training each binary classification model, and the behavioral features of the user corresponding to the target classification feature are used when training each binary classification model; Training a label model according to the target classification feature of each user in the second target user group set and the first probability that the label is each target classification feature; The second probability that the label of the user to be identified is each target classification feature is determined according to each of the two classification models, and the label of the user to be identified is generated according to the second probability and the label model.

2. The method according to claim 1, characterized in that Generating a tag of the user to be identified according to the second probability and the tag model includes: The second probability that the label of the user to be identified is each target classification feature is input into the label model to obtain the third probability that the label of the user to be identified is each target classification feature; according to the third probability that the label of the user to be identified is each target classification feature, the target classification feature or target classification feature list of the user to be identified is determined.

3. The method according to claim 1, characterized in that The training features of the binary classification model are determined according to the target classification features.

4. The method according to claim 1, characterized in that The classification feature is a shopping channel; the user's behavior feature includes at least one of the following: browsing items, adding items to a shopping cart, collecting items, and following items.

5. A device for generating user tags, characterized in that: include: A user group division module, which divides all users into multiple user groups according to user classification characteristics; among all users, users who meet the target classification characteristics are the first target user group; and users who only meet the target classification characteristics are the second target user group; A binary classification model training module, for each of the target classification features, using the corresponding first target user group as a positive sample and the remaining users among all users as negative samples to train the binary classification model; determining a first probability that the label of each user in the second target user group set is the target classification feature according to the behavioral features of each user in the second target user group set and the binary classification model; when the number of the target classification features is two or more, different training features are used when training each binary classification model, and the behavioral features of the user corresponding to the target classification feature are used when training each binary classification model; A label model training module, which trains a label model according to the target classification feature of each user in the second target user group set and the first probability that the label is each target classification feature; The label generation module determines the second probability that the label of the user to be identified is each target classification feature according to each of the two classification models, and generates the label of the user to be identified according to the second probability and the label model.

6. The device according to claim 5, characterized in that The tag generation module generates a tag of a user to be identified according to the second probability and the tag model, including: The second probability that the label of the user to be identified is each target classification feature is input into the label model to obtain the third probability that the label of the user to be identified is each target classification feature; according to the third probability that the label of the user to be identified is each target classification feature, the target classification feature or target classification feature list of the user to be identified is determined.

7. The device according to claim 5, characterized in that The training features of the binary classification model are determined according to the target classification features.

8. The device according to claim 5, characterized in that The classification feature is a shopping channel; the user's behavior feature includes at least one of the following: browsing items, adding items to a shopping cart, collecting items, and following items.

9. An electronic device for generating user tags, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 4.

10. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images