User portrait label generation method and device, storage medium and program product

Abstract

The embodiment of the invention provides a user portrait tag generation method and device, a storage medium and a program product, and the method comprises the steps: obtaining target data of a user, the target data comprises operation behavior data of the user for a target page and image data of a product image in the target page, and generating a user portrait tag based on a preset feature index; the target data are preprocessed, feature data corresponding to preset feature indexes are obtained, the preset feature indexes comprise a plurality of first-order feature indexes and at least one second-order feature index, the at least one second-order feature index is determined according to the at least one first-order feature index, and the feature data are input into a target model, and obtaining a user portrait label of the user, wherein the target model is obtained based on preset feature index training. According to the method provided by the embodiment of the invention, the comprehensiveness of the user portrait label and the efficiency of determining the user portrait label are improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of artificial intelligence technology, and in particular, to a method, device, storage medium and program product for generating a user portrait label. Background Art

[0002] With the rapid development of Internet technology, e-commerce has become an indispensable part of the global economy, greatly changing people's shopping methods and business models. Different users have different preferences and habits, and setting portrait tags for users is one of the key technologies in the development of e-commerce.

[0003] In the related technology, taking the scenario of consumers purchasing goods as an example, with the consumer's consent, the consumer's online browsing behavior can be recorded, so that sales personnel can set portrait labels for consumers (such as high consumption level, brand preference, marketing sensitivity, etc.) by viewing the consumer's online browsing behavior, artificially infer the consumer's purchasing intention, screen the consumer, and further contact the consumer to determine the purchase demand, etc.

[0004] However, in the process of implementing this application, the inventors found that there are at least the following problems in the prior art: in the above method, the efficiency of determining user portrait labels is low and the comprehensiveness is poor. Summary of the invention

[0005] The embodiments of the present application provide a method, device, storage medium and program product for generating user portrait labels to improve the efficiency of determining user portrait labels and the comprehensiveness of user portrait labels.

[0006] In a first aspect, an embodiment of the present application provides a method for generating a user portrait tag, comprising:

[0007] Acquire target data of the user, the target data including: operation behavior data of the user on the target page and image data of the product image in the target page;

[0008] Based on preset characteristic indicators, the target data is preprocessed to obtain characteristic data corresponding to the preset characteristic indicators; the preset characteristic indicators include multiple first-order characteristic indicators and at least one second-order characteristic indicator, and the at least one second-order characteristic indicator is determined according to at least one first-order characteristic indicator;

[0009] The feature data is input into a target model to obtain a user portrait label of the user; the target model is trained based on the preset feature indicators.

[0010] In a possible design, before inputting the feature data into the target model to obtain the user portrait label of the user, the method further includes:

[0011] Performing missing data filling, standardization or normalization processing on the feature data to obtain pre-processed feature data;

[0012] The step of inputting the feature data into a target model to obtain a user portrait label of the user includes:

[0013] The preprocessed feature data is input into a target model to obtain a user portrait label of the user.

[0014] In a possible design, after inputting the feature data into the target model to obtain the user portrait label of the user, the method further includes:

[0015] If the user portrait tag of the user meets the preset conditions, manually verify the user portrait tag of the user to obtain a verified user portrait tag;

[0016] Storing the verified user portrait tag and the user's feature data in a preset storage space;

[0017] When determining a user portrait label for a new user, the feature data of the new user is compared with the feature data of the user stored in the preset storage space to obtain a similarity value;

[0018] If the similarity value is greater than a preset value, the user portrait tag of the user is determined as the user portrait tag of the new user.

[0019] In a possible design, the user is a product purchaser; the target page is a product display page; the multiple first-order indicators include at least one of the following: total purchase amount, total purchase times, category of purchased goods, number of additions to shopping carts, number of entries into a product details page, number of transactions within a unit period, zoom ratio of the product image, style of the product image, hue of the product image, and main body area ratio of the product image; the main body area ratio refers to the ratio of the area occupied by the product main body in the product image to the total area of ​​the product image.

[0020] In one possible design, the user is a product seller; the target page includes a product poster selection page; the product poster selection page includes multiple poster images corresponding to the target product, and the multiple first-order indicators include at least one of the following: the total number of selected poster images, the number of views of the poster image, the viewing time of the poster image, the style of the poster image, the tone of the poster image, and the proportion of the main area of ​​the poster image.

[0021] In a possible design, before inputting the feature data into the target model, the process further includes:

[0022] Acquire a sample set; the sample set includes target sample data of multiple sample users; the target sample data includes: operation behavior data for a target page and image data of a product image in the target page;

[0023] Determine a plurality of candidate feature indicators according to the sample set; the candidate feature indicators include a plurality of candidate first-order feature indicators and a plurality of candidate second-order feature indicators;

[0024] Screening the plurality of candidate feature indicators based on feature engineering to obtain preset feature indicators;

[0025] Based on the preset characteristic index, the sample set is processed to obtain sample characteristic data corresponding to the preset characteristic index;

[0026] The initial model is trained according to the sample feature data to obtain the target model.

[0027] In a possible design, determining a plurality of candidate feature indicators according to the sample set includes:

[0028] Determine a plurality of candidate first-order feature indicators according to the sample set;

[0029] A plurality of candidate first-order feature indicators are multiplied in pairs or a plurality of candidate first-order feature indicators are squared to obtain a plurality of candidate second-order feature indicators.

[0030] In a second aspect, an embodiment of the present application provides a device for generating a user portrait label, including:

[0031] An acquisition module, used to acquire target data of a user, wherein the target data includes: operation behavior data of the user on a target page and image data of a product image in the target page;

[0032] A preprocessing module, used to preprocess the target data based on preset characteristic indicators to obtain characteristic data corresponding to the preset characteristic indicators; the preset characteristic indicators include multiple first-order characteristic indicators and at least one second-order characteristic indicator, and the at least one second-order characteristic indicator is determined based on at least one first-order characteristic indicator;

[0033] An input module is used to input the feature data into a target model to obtain a user portrait label of the user; the target model is trained based on the preset feature indicators.

[0034] In a third aspect, an embodiment of the present application provides a device for generating a user portrait label, including: at least one processor and a memory;

[0035] The memory stores computer-executable instructions;

[0036] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method described in the first aspect and various possible designs of the first aspect.

[0037] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the method described in the first aspect and various possible designs of the first aspect are implemented.

[0038] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the method described in the first aspect and various possible designs of the first aspect.

[0039] The method, device, storage medium and program product for generating user portrait labels provided in the present embodiment include obtaining the user's target data, the target data include: the user's operation behavior data for the target page and the image data of the product image in the target page, preprocessing the target data based on preset feature indicators, obtaining feature data corresponding to the preset feature indicators, the preset feature indicators include multiple first-order feature indicators and at least one second-order feature indicator, at least one second-order feature indicator is determined based on at least one first-order feature indicator, inputting the feature data into the target model, obtaining the user portrait label of the user, and the target model is obtained by training based on the preset feature indicators. The method provided in the embodiment of the present application, by fusing the user's operation behavior data and image data, and determining the feature data based on the first-order feature indicators and the second-order feature indicators, makes the feature data contain more comprehensive information, and then automatically classifies the feature data through the target model, and can quickly obtain the user portrait label, thereby improving the efficiency and comprehensiveness of the user portrait label. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0041] Figure 1 A schematic diagram of an application scenario of a method for generating a user portrait label provided in an embodiment of the present application;

[0042] Figure 2 Schematic diagram of the process of generating a user portrait label provided in the embodiment of the present application Figure 1 ;

[0043] Figure 3 Schematic diagram of the process of generating a user portrait label provided in the embodiment of the present application Figure 2 ;

[0044] Figure 4 A flowchart of a method for training a target model provided in an embodiment of the present application;

[0045] Figure 5 A schematic diagram of the structure of a device for generating user portrait labels provided in an embodiment of the present application;

[0046] Figure 6 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0048] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0049] It should be noted that the method, device, storage medium and program product for generating user portrait labels provided in this application can be used in the field of artificial intelligence technology, and can also be used in any field outside the field of artificial intelligence technology. The application field of the method, device, storage medium and program product for generating user portrait labels provided in this application is not limited.

[0050] With the rapid development of Internet technology, e-commerce has become an indispensable part of the global economy, greatly changing people's shopping methods and business models. Different users have different preferences and habits, and setting portrait tags for users is one of the key technologies in the development of e-commerce.

[0051] In the related technology, taking the scenario of consumers purchasing goods as an example, with the consumer's consent, the consumer's online browsing behavior can be recorded, so that sales personnel can set portrait labels for consumers (such as high consumption level, brand preference, marketing sensitivity, etc.) by viewing the consumer's online browsing behavior, artificially infer the consumer's purchasing intention, screen the consumer, and further contact the consumer to determine the purchase demand, etc.

[0052] However, in the above method, the efficiency of determining user portrait labels is low and the comprehensiveness is poor.

[0053] In order to solve the above technical problems, the inventors of this application have found that the user portrait label can be automatically determined through the target model to improve the determination efficiency, and the user portrait label can be determined by combining image features and user operation behavior data to improve the comprehensiveness of the user portrait label. Based on this, the embodiment of this application provides a method for generating a user portrait label.

[0054] Figure 1 A schematic diagram of an application scenario of a method for generating a user portrait tag provided in an embodiment of the present application. Figure 1 As shown, the terminal device 101 is in communication connection with the server 102. The terminal device 101 is installed with a client application.

[0055] In the specific implementation process, after the client application of the terminal device 101 is started, the user's touch operation on the target page is received. In this process, the terminal device 101 collects the user's operation behavior data and the target data such as the image data of the product image in the target page, and sends the target data to the server 102. The server 102 obtains the user's target data, and the target data includes: the user's operation behavior data for the target page and the image data of the product image in the target page. Based on the preset feature index, the target data is preprocessed to obtain the feature data corresponding to the preset feature index. The preset feature index includes multiple first-order feature indexes and at least one second-order feature index. At least one second-order feature index is determined based on at least one first-order feature index. The feature data is input into the target model to obtain the user's user portrait label. The target model is obtained by training based on the preset feature index. The method for generating a user portrait label provided in the embodiment of the present application, by fusing the user's operation behavior data and image data, and determining the feature data based on the first-order feature index and the second-order feature index, makes the feature data contain more comprehensive information, and then automatically classifies the feature data through the target model, and can quickly obtain the user portrait label, thereby improving the efficiency and comprehensiveness of the user portrait label.

[0056] It should be noted that Figure 1The scenario diagram shown is only an example. The method and scenario for generating user portrait labels described in the embodiment of the present application are intended to more clearly illustrate the technical solution of the embodiment of the present application, and do not constitute a limitation on the technical solution provided in the embodiment of the present application. A person of ordinary skill in the art can appreciate that with the evolution of the system and the emergence of new business scenarios, the technical solution provided in the embodiment of the present application is equally applicable to similar technical problems.

[0057] The technical solution of the present application is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0058] Figure 2 Schematic diagram of the process of generating a user portrait label provided in the embodiment of the present application Figure 1 .

[0059] like Figure 2 As shown, the method includes:

[0060] 201. Obtain target data of a user, where the target data includes: operation behavior data of the user on a target page and image data of a product image in the target page.

[0061] The execution subject of this embodiment may be a terminal device or a server.

[0062] In this embodiment, the target page may be a page of a client application, and the client may be a seller client (ie, a merchant client) or a consumer client.

[0063] Specifically, taking the target page as the product display page of the buyer's client as an example, the user's click operation, zooming and other operation behavior data on the target page can be obtained. The target page can include multiple product cards. When the card is touched, the user can enter the product details page. The card can include the product image of the product, and may also include the product's function introduction, price information, etc.

[0064] In some embodiments, tracking technology can be used to obtain user operation behavior data. Specifically, browsing behavior tracking can record information such as the pages browsed by the user and the duration of stay; click behavior tracking can record the user's click events, such as clicking on product details, adding to the shopping cart, etc.; heartbeat tracking can be used to detect whether the user is still active on the page.

[0065] 202. Preprocess the target data based on preset characteristic indicators to obtain characteristic data corresponding to the preset characteristic indicators; the preset characteristic indicators include multiple first-order characteristic indicators and at least one second-order characteristic indicator, and the at least one second-order characteristic indicator is determined based on at least one first-order characteristic indicator.

[0066] Specifically, after obtaining the target data of the user, the target data may be statistically calculated and processed according to the characteristic indicators adopted by the target model to obtain characteristic data.

[0067] For example, the number of purchases can be counted to obtain the characteristic value of the total number of purchases, the number of times added to the shopping cart can be counted to obtain the characteristics of the number of times added to the shopping cart, the number of times the product details page is entered can be counted to obtain the characteristic value of the number of times the product is browsed, and further, the characteristic values ​​of the number of times the product is browsed after the last purchase, the number of times the product is added to the shopping cart after the last purchase, and the number of times the product is searched after the last purchase can be specifically counted. The characteristic values ​​of each characteristic indicator calculated are the characteristic data.

[0068] In this embodiment, the process of determining the second-order index according to the first-order index can be described with reference to the following embodiment, which will not be repeated here.

[0069] In some embodiments, in order to reduce the feature magnitude differences between feature values ​​of different feature indicators, normalization and other preprocessing may be performed before the feature data is input into the target model. Specifically, before inputting the feature data into the target model and obtaining the user portrait label of the user, it may also include: filling in missing data and standardizing or normalizing the feature data to obtain the preprocessed feature data; correspondingly, inputting the feature data into the target model and obtaining the user portrait label of the user may include: inputting the preprocessed feature data into the target model to obtain the user portrait label of the user.

[0070] Specifically, missing data can be filled in the following ways: First, mean filling: use the average value of the column where the missing data is located for filling. This method is simple, but may introduce bias, especially when there are many missing values. Second, median filling: use the median of the column where the missing data is located for filling. This method is more robust for skewed distributed data. Third, interpolation: interpolate according to the data points before and after the missing value, such as linear interpolation, polynomial interpolation, etc. This method is suitable for time series data or data with obvious trends. Fourth, model prediction: use machine learning models (such as regression models) to predict missing values. This method is more complicated, but it can consider more data features and may obtain more accurate filling results. By filling in missing data, the data set is complete for subsequent analysis and modeling.

[0071] During the standardization or normalization process, if the data distribution is close to normal distribution, or the algorithm is sensitive to the scale of the data (such as many machine learning algorithms), standardization is used. If the data has no obvious distribution characteristics, or the algorithm is sensitive to the data range (such as the activation function in the neural network), normalization is used. It can reduce the impact of feature magnitude differences, so that features of different magnitudes can be compared and calculated at the same scale, and can improve the stability and performance of the model. In particular, when using optimization algorithms based on gradient descent, standardization can accelerate convergence.

[0072] 203. Input the feature data into a target model to obtain a user portrait label of the user; the target model is trained based on the preset feature indicators.

[0073] Specifically, after determining the feature data, the feature data can be input into the target model, and the feature data can be processed by the target model to obtain multiple corresponding user portrait labels. User portrait labels represent user preferences and habits. A single user can have multiple user portrait labels, such as: youth group, health insurance concern, high income, etc.

[0074] The method for generating user portrait labels provided in this embodiment integrates the user's operation behavior data and image data, and determines the feature data based on the first-order feature indicators and the second-order feature indicators, so that the feature data contains more comprehensive information, and then automatically classifies the feature data through the target model, so as to quickly obtain the user portrait labels, thereby improving the efficiency and comprehensiveness of the user portrait labels.

[0075] The method for generating user portrait labels provided in the embodiment of the present application can be applied to different scenarios. One is to use the relevant data of consumers visiting online stores as target data to determine the user portrait labels of consumers. Another method can be that when sales staff adjust the display of goods in their own shops, they can select a poster from multiple candidate posters for each product to display on the display page of their own shop for users to purchase. The candidate posters can be poster images with different styles, tones and different main body area ratios specially designed for the product by a professional design team. By adopting the method of the present application, the user portrait labels of sales staff can be determined so that the platform can push candidate posters based on the preferences of sales staff, and it is beneficial for professional design teams to improve design ideas.

[0076] In some embodiments, the user is a product purchaser; the target page is a product display page; the multiple first-order indicators include at least one of the following: total purchase amount, total purchase times, category of purchased goods, number of times added to shopping cart, number of times entering product details page, number of transactions within a unit period, zoom ratio of product image, style of product image, hue of product image, main body area ratio of product image; the main body area ratio refers to the ratio between the area occupied by the main body of the product in the product image and the total area of ​​the product image. Among them, the function of obtaining the zoom ratio in real time can usually be realized in e-commerce and other applications. The larger the user's zoom ratio of the picture, the greater the user's interest in the picture. In addition, everyone has different aesthetics, and their preferences for the style, hue, main body area ratio, etc. of product images may be different.

[0077] In some embodiments, the user is a product seller; the target page includes a product poster selection page; the product poster selection page includes multiple poster images corresponding to the target product, and the multiple first-order indicators include at least one of the following: the total number of selected poster images, the number of views of the poster image, the viewing time of the poster image, the style of the poster image, the tone of the poster image, and the proportion of the main area of ​​the poster image.

[0078] For example, there are a variety of insurance products in e-commerce applications such as online shopping malls, and each insurance product has promotional posters with different styles, tones, and main body areas. Based on the customer tags formed by image features, the platform can filter out posters that match the corresponding customer tags for customers with corresponding customer tags of style, tones, and main body area. Sales staff can actively share posters that meet customer preferences with customers based on the platform's screening results. In addition, when customers browse products on the platform by themselves, they can also see the product promotion posters they like.

[0079] Figure 3 Schematic diagram of the process of generating a user portrait label provided in the embodiment of the present application Figure 2 .like Figure 3 As shown, based on the above embodiment, for example Figure 2 Based on the embodiment shown, this embodiment describes in detail the process of manually verifying the results after the target model outputs the results and making full use of the manual verification results to improve the accuracy. The method includes:

[0080] 301. Obtain target data of a user, where the target data includes: operation behavior data of the user on a target page and image data of a product image in the target page.

[0081] 302. Preprocess the target data based on preset characteristic indicators to obtain characteristic data corresponding to the preset characteristic indicators; the preset characteristic indicators include multiple first-order characteristic indicators and at least one second-order characteristic indicator, and the at least one second-order characteristic indicator is determined based on at least one first-order characteristic indicator.

[0082] 303. Input the feature data into a target model to obtain a user portrait label of the user; the target model is trained based on the preset feature indicators.

[0083] Steps 301 to 303 in this embodiment are similar to steps 201 to 203 in the above embodiment, and are not described again here.

[0084] 304. If the user portrait tag of the user meets the preset conditions, the user portrait tag of the user is manually verified to obtain a verified user portrait tag; the verified user portrait tag and the feature data of the user are stored in a preset storage space.

[0085] 305. When determining a user portrait label for a new user, compare the feature data of the new user with the feature data of the user stored in the preset storage space to obtain a similarity value; if the similarity value is greater than the preset value, determine the user portrait label of the user as the user portrait label of the new user.

[0086] Specifically, after obtaining the user portrait label of the user through the target model, the output user portrait label can be manually verified to ensure the accuracy of the output result. Of course, in order to improve efficiency and save manpower, only the output results that meet the preset conditions can be manually verified. The preset condition can be that the probability value of the model output result is lower than the preset probability value, that is, when the confidence of the model output result is low. After manual verification, the relevant data of the user (such as user portrait label, target data, feature data, etc.) can be stored in a preset storage space. So that the next time a similar user is encountered, the similarity between the feature data of the current user and the feature data of the stored user can be compared. When the similarity value is greater than the preset value, it indicates that the current user and the user participating in the comparison have the same user portrait label, so the user portrait label of the user participating in the comparison can be directly determined as the user portrait label of the current user. There is no need to predict through the target model, which can improve the accuracy of the user portrait label output by the model, and at the same time make full use of the manually verified user portrait label.

[0087] The method for generating user portrait labels provided in this embodiment manually verifies the model output results, and stores the manually verified user portrait labels and related feature data or target data, so as to facilitate the subsequent direct comparison of the stored related data (feature data or target data) with the related data of the current user (the user whose user portrait label currently needs to be inferred), quickly determine the user portrait label of the current user, and improve accuracy.

[0088] Figure 4 A flow chart of a method for training a target model provided in an embodiment of the present application. Figure 4 As shown, the method includes:

[0089] 401. Obtain a sample set; the sample set includes target sample data of multiple sample users; the target sample data includes: operation behavior data for a target page and image data of a product image in the target page.

[0090] 402. Determine a plurality of candidate feature indicators according to the sample set; the candidate feature indicators include a plurality of candidate first-order feature indicators and a plurality of candidate second-order feature indicators.

[0091] Specifically, we can integrate the data of users from multiple channels, not limited to a single online store, aggregate product data and user behavior data, and extract as many first-order candidate feature indicators as possible that can represent user preferences, such as:

[0092] Transaction-related characteristic indicators may include total purchase times, total purchase amount, average purchase amount, number of transactions within a unit period, etc.

[0093] Product-related characteristic indicators may include the quantity of different products purchased, the quantity of different product categories purchased, etc.

[0094] Behavior-related characteristic indicators may include the total number of times a product is viewed, the number of different products viewed, the number of times products are added to the shopping cart, etc.

[0095] User information-related characteristic indicators may include age, gender, and number of days registered on the platform.

[0096] Interaction feature indicators may include the number of times a product was browsed after the last purchase, the number of times a product was added to the shopping cart after the last purchase, the number of times a product was searched after the last purchase, etc.

[0097] Image feature indicators may include image tone, image style (realistic style, cartoon, abstract style, romanticism, etc.), image main body area ratio, etc.

[0098] In some embodiments, image features may include the length of time a user stays on each picture, whether to zoom in or out, and whether to recognize a QR code in the picture when the user browses the product display pictures in the online mall. It may also include the color tone, style (realistic style, cartoon, abstract style, romanticism, etc.), main body area, etc. of the picture. In the specific implementation process, pictures can be screened based on the length of stay, zooming action, etc., and text recognition can be performed on the pictures to extract keywords to understand which aspects of the content introduction the user is concerned about. (For example, a series of pictures may be included for a product, and different pictures introduce different aspects of the content, such as product features, product uses, product prices, etc. Assuming that the user is concerned about product features, then in subsequent recommendations, more detailed pictures of product features can be provided to the user, or product features can be used as cover pictures, etc.).

[0099] In some embodiments, determining multiple candidate feature indicators based on the sample set may include: determining multiple candidate first-order feature indicators based on the sample set; multiplying the multiple candidate first-order feature indicators in pairs or squaring the multiple candidate first-order feature indicators to obtain multiple candidate second-order feature indicators.

[0100] Exemplarily, for example, assume that 30 first-order candidate indicators are extracted based on the target sample data of user A; then each first-order candidate indicator can be squared to obtain 30 second-order candidate indicators, and each first-order candidate indicator can be multiplied two by two to obtain second-order candidate indicators, obtaining C(n,2)=n(n-1) / 2, that is, 435 second-order candidate indicators, then model training can be performed based on all or part of the 30 first-order candidate indicators and / or 465 second-order candidate indicators.

[0101] 403. Screen the multiple candidate feature indicators based on feature engineering to obtain preset feature indicators.

[0102] Specifically, feature engineering is an important technology in machine learning. It enables machine learning algorithms to better use data features for model training and prediction by preprocessing, transforming and extracting data. In the algorithm, the indicators used are screened by feature engineering, and it is a crucial step to select the most useful features for prediction, reduce redundancy and improve data quality. This can be achieved by using indicator correlation analysis, feature importance ranking and other methods. Among them, indicator correlation analysis is to determine the importance of features by calculating the correlation coefficient between features and target variables. The range of the correlation coefficient is [-1,1], and the larger the absolute value, the stronger the correlation. In feature engineering, the correlation coefficient between each feature and the target variable can be calculated, and those features that are highly correlated with the target variable can be selected for retention, while those features that are less correlated or irrelevant to the target variable can be eliminated. Feature importance ranking refers to sorting features to determine their degree of influence on the target variable. It can be achieved through a variety of machine learning algorithms, such as decision trees, random forests, gradient boosting, etc. For example, decision tree algorithm: Use decision tree algorithms (such as CART algorithm) to build models, and evaluate the importance of features based on the number of node splits or information gain of features in the decision tree.

[0103] 404. Based on the preset characteristic index, the sample set is processed to obtain sample characteristic data corresponding to the preset characteristic index.

[0104] Specifically, after the first-order characteristic index and the second-order characteristic index are determined, characteristic values ​​corresponding to different characteristic indexes can be extracted from the data of the sample set based on the determined characteristic indexes, thereby obtaining sample characteristic data.

[0105] 405. Train the initial model according to the sample feature data to obtain the target model.

[0106] Specifically, multiple samples correspond to multiple users one by one. Multiple samples can be input into the initial model in batches, and the initial model is trained to obtain the target model.

[0107] In this embodiment, the initial model may be a machine learning model, such as a multi-label support vector machine algorithm model.

[0108] The target model training method provided in this embodiment integrates the user's operation behavior data and image data, and determines the sample feature data based on the first-order feature indicators and the second-order feature indicators, so that the sample feature data contains more comprehensive information, and then the target model is trained based on the sample feature data, so that the user portrait label predicted by the target model is more comprehensive, accurate and more efficient.

[0109] Figure 5This is a schematic diagram of the structure of a device for generating user portrait labels provided in an embodiment of the present application. Figure 5 As shown, the user portrait label generation device 50 includes: an acquisition module 501, a preprocessing module 502 and an input module 503.

[0110] The acquisition module 501 is used to acquire the user's target data, wherein the target data includes: the user's operation behavior data on the target page and the image data of the product image in the target page.

[0111] The preprocessing module 502 is used to preprocess the target data based on preset feature indicators to obtain feature data corresponding to the preset feature indicators; the preset feature indicators include multiple first-order feature indicators and at least one second-order feature indicator, and the at least one second-order feature indicator is determined based on at least one of the first-order feature indicators.

[0112] The input module 503 is used to input the feature data into a target model to obtain a user portrait label of the user; the target model is trained based on the preset feature indicators.

[0113] The user portrait label generation device provided in the embodiment of the present application integrates the user's operation behavior data and image data, and determines the feature data based on the first-order feature indicators and the second-order feature indicators, so that the feature data contains more comprehensive information, and then automatically classifies the feature data through the target model, so as to quickly obtain the user portrait label, thereby improving the efficiency and comprehensiveness of the user portrait label.

[0114] In some embodiments, the preprocessing module 502 is also used to: fill in missing data, and standardize or normalize the feature data to obtain preprocessed feature data; correspondingly, the input module 503 is specifically used to: input the preprocessed feature data into the target model to obtain the user portrait label of the user.

[0115] In some embodiments, the input module 503 is also used for: if the user portrait tag of the user meets the preset conditions, manually verifying the user portrait tag of the user to obtain the verified user portrait tag; storing the verified user portrait tag and the characteristic data of the user in a preset storage space; when determining the user portrait tag for a new user, comparing the characteristic data of the new user with the characteristic data of the user stored in the preset storage space to obtain a similarity value; if the similarity value is greater than the preset value, determining the user portrait tag of the user as the user portrait tag of the new user.

[0116] In some embodiments, the user is a product purchaser; the target page is a product display page; the multiple first-order indicators include at least one of the following: total purchase amount, total purchase times, category of purchased goods, number of additions to shopping carts, number of entries into product details page, number of transactions within a unit period, zoom ratio of product image, style of product image, hue of product image, and main body area ratio of product image; the main body area ratio refers to the ratio between the area occupied by the product main body in the product image and the total area of ​​the product image.

[0117] In some embodiments, the user is a product seller; the target page includes a product poster selection page; the product poster selection page includes multiple poster images corresponding to the target product, and the multiple first-order indicators include at least one of the following: the total number of selected poster images, the number of views of the poster image, the viewing time of the poster image, the style of the poster image, the tone of the poster image, and the proportion of the main area of ​​the poster image.

[0118] In some embodiments, the device 50 also includes a training module (not shown), which is used to: obtain a sample set; the sample set includes target sample data of multiple sample users; the target sample data includes: operation behavior data for a target page and image data of a product image in the target page; determine multiple candidate feature indicators based on the sample set; the candidate feature indicators include multiple candidate first-order feature indicators and multiple candidate second-order feature indicators; screen the multiple candidate feature indicators based on feature engineering to obtain preset feature indicators; based on the preset feature indicators, process the sample set to obtain sample feature data corresponding to the preset feature indicators; train the initial model based on the sample feature data to obtain the target model.

[0119] In some embodiments, the training module is specifically used to: determine multiple candidate first-order feature indicators based on the sample set; multiply the multiple candidate first-order feature indicators by two or square the multiple candidate first-order feature indicators to obtain multiple candidate second-order feature indicators.

[0120] The user portrait label generation device provided in the embodiment of the present application can be used to execute the above-mentioned method embodiment. Its implementation principle and technical effects are similar, and this embodiment will not be repeated here.

[0121] Figure 6 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application, which device may be a computer, a tablet device, a server, etc.

[0122] The device 60 may include one or more of the following components: a processing component 601 , a memory 602 , a power component 603 , a multimedia component 604 , an audio component 605 , an input / output (I / O) interface 606 , a sensor component 607 , and a communication component 608 .

[0123] The processing component 601 generally controls the overall operation of the device 60, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 601 may include one or more processors 609 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 601 may include one or more modules to facilitate the interaction between the processing component 601 and other components. For example, the processing component 601 may include a multimedia module to facilitate the interaction between the multimedia component 604 and the processing component 601.

[0124] The memory 602 is configured to store various types of data to support operations on the device 60. Examples of such data include instructions for any application or method operating on the device 60, contact data, phone book data, messages, pictures, videos, etc. The memory 602 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0125] The power supply component 603 provides power to the various components of the device 60. The power supply component 603 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device 60.

[0126] The multimedia component 604 includes a screen that provides an output interface between the device 60 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 604 includes a front camera and / or a rear camera. When the device 60 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and the rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0127] The audio component 605 is configured to output and / or input audio signals. For example, the audio component 605 includes a microphone (MIC), and when the device 60 is in an operation mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 602 or sent via the communication component 608. In some embodiments, the audio component 605 also includes a speaker for outputting audio signals.

[0128] I / O interface 606 provides an interface between processing component 601 and peripheral interface modules, which may be keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: home button, volume button, start button, and lock button.

[0129] The sensor assembly 607 includes one or more sensors for providing various aspects of status assessment for the device 60. For example, the sensor assembly 607 can detect the open / closed state of the device 60, the relative positioning of components, such as the display and keypad of the device 60, and the sensor assembly 607 can also detect the position change of the device 60 or a component of the device 60, the presence or absence of user contact with the device 60, the orientation or acceleration / deceleration of the device 60, and the temperature change of the device 60. The sensor assembly 607 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 607 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 607 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0130] The communication component 608 is configured to facilitate wired or wireless communication between the device 60 and other devices. The device 60 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 608 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 608 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0131] In an exemplary embodiment, the device 60 may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above methods.

[0132] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 602 including instructions, and the instructions can be executed by a processor 609 of the device 60 to perform the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0133] The computer-readable storage medium mentioned above can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0134] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0135] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0136] An embodiment of the present application also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the method for generating a user portrait label performed by the user portrait label generation device as described above.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for generating a user portrait label, characterized in that: include: Acquire target data of the user, the target data including: operation behavior data of the user on the target page and image data of the product image in the target page; Based on preset characteristic indicators, the target data is preprocessed to obtain characteristic data corresponding to the preset characteristic indicators; the preset characteristic indicators include multiple first-order characteristic indicators and at least one second-order characteristic indicator, and the at least one second-order characteristic indicator is determined according to at least one first-order characteristic indicator; The feature data is input into a target model to obtain a user portrait label of the user; the target model is trained based on the preset feature indicators.

2. The method according to claim 1, characterized in that Before inputting the feature data into the target model to obtain the user portrait label of the user, the method further includes: Performing missing data filling, standardization or normalization processing on the feature data to obtain pre-processed feature data; The step of inputting the feature data into a target model to obtain a user portrait label of the user includes: The preprocessed feature data is input into a target model to obtain a user portrait label of the user.

3. The method according to claim 1, characterized in that After inputting the feature data into the target model to obtain the user portrait label of the user, the method further includes: If the user portrait tag of the user meets the preset conditions, manually verify the user portrait tag of the user to obtain a verified user portrait tag; Storing the verified user portrait tag and the user's feature data in a preset storage space; When determining a user portrait label for a new user, the feature data of the new user is compared with the feature data of the user stored in the preset storage space to obtain a similarity value; If the similarity value is greater than a preset value, the user portrait tag of the user is determined as the user portrait tag of the new user.

4. The method according to claim 1, characterized in that: The user is a product purchaser; the target page is a product display page; the multiple first-order indicators include at least one of the following: total purchase amount, total purchase times, category of purchased goods, number of additions to shopping carts, number of entries into product details page, number of transactions within a unit period, zoom ratio of product image, style of product image, hue of product image, and main body area ratio of product image; the main body area ratio refers to the ratio between the area occupied by the product main body in the product image and the total area of ​​the product image.

5. The method according to claim 1, characterized in that The user is a product seller; the target page includes a product poster selection page; the product poster selection page includes multiple poster images corresponding to the target product, and the multiple first-order indicators include at least one of the following: the total number of selected poster images, the number of views of the poster image, the viewing time of the poster image, the style of the poster image, the color tone of the poster image, and the proportion of the main area of ​​the poster image.

6. The method according to any one of claims 1 to 5, characterized in that: Before inputting the feature data into the target model, the method further includes: Acquire a sample set; the sample set includes target sample data of multiple sample users; the target sample data includes: operation behavior data for a target page and image data of a product image in the target page; Determine a plurality of candidate feature indicators according to the sample set; the candidate feature indicators include a plurality of candidate first-order feature indicators and a plurality of candidate second-order feature indicators; Screening the plurality of candidate feature indicators based on feature engineering to obtain preset feature indicators; Based on the preset characteristic index, the sample set is processed to obtain sample characteristic data corresponding to the preset characteristic index; The initial model is trained according to the sample feature data to obtain the target model.

7. The method according to claim 6, characterized in that The determining of a plurality of candidate feature indicators according to the sample set includes: Determine a plurality of candidate first-order feature indicators according to the sample set; A plurality of candidate first-order feature indicators are multiplied in pairs or a plurality of candidate first-order feature indicators are squared to obtain a plurality of candidate second-order feature indicators.

8. A device for generating a user portrait label, characterized in that: include: An acquisition module, used to acquire target data of a user, wherein the target data includes: operation behavior data of the user on a target page and image data of a product image in the target page; A preprocessing module, used to preprocess the target data based on preset characteristic indicators to obtain characteristic data corresponding to the preset characteristic indicators; the preset characteristic indicators include multiple first-order characteristic indicators and at least one second-order characteristic indicator, and the at least one second-order characteristic indicator is determined based on at least one first-order characteristic indicator; An input module is used to input the feature data into a target model to obtain a user portrait label of the user; the target model is trained based on the preset feature indicators.

9. A device for generating a user portrait label, characterized in that: include: at least one processor and memory; The memory stores computer-executable instructions; The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method for generating a user portrait label as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions. When the processor executes the computer-executable instructions, the method for generating a user portrait label as described in any one of claims 1 to 7 is implemented.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for generating a user portrait label according to any one of claims 1 to 7 is implemented.

Images