A method, system and medium for screening target users

Abstract

The embodiments of the present application provide a method, system and medium for screening target users. In the method, a screening strategy analysis is first performed based on the target marketing scenario and the target product attribute information to determine the target user screening strategy. Subsequently, based on a preset natural language model, the target user screening strategy is vector encoded to obtain a target product coding vector. Furthermore, similarity screening is performed based on the target product coding vector and a preset user vector database to determine multiple groups of first user coding vectors that match the target product coding vector from the preset user vector database. Finally, the target user is determined based on each first user coding vector and the target product coding vector. In this way, the product characteristics of the target product and the shopping behavior preferences of different users can be more comprehensively characterized, thereby improving the screening accuracy of the target user.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a target user screening method, system and medium. Background Art

[0002] In the current scenario-based marketing plan, for different types of products, marketers need to screen potential target users who are willing to learn about the products based on their own experience. The accuracy of screening target users is completely determined by the marketers' experience, and the accuracy of screening target users is relatively poor. Summary of the invention

[0003] Based on the above problems, in order to improve the accuracy of screening for target users, the embodiments of the present application provide a method, system and medium for screening target users.

[0004] The embodiments of the present application disclose the following technical solutions:

[0005] In a first aspect, an embodiment of the present application provides a method for screening target users, comprising:

[0006] Conduct screening strategy analysis based on target marketing scenarios and target product attribute information to determine target user screening strategies;

[0007] Based on a preset natural language model, the target user screening strategy is vector-encoded to obtain a target product coding vector; the target product coding vector is used to characterize the product features of the target product and the marketing preference of the target marketing scenario;

[0008] Performing similarity screening based on the target product coding vector and a preset user vector database to determine multiple groups of first user coding vectors matching the target product coding vector from the preset user vector database; the first user coding vector is used to characterize the historical shopping behavior characteristics of the user;

[0009] A target user is determined according to each of the first user coding vectors and the target product coding vector.

[0010] In a possible implementation, the method further includes:

[0011] Performing data standardization processing on the historical shopping logs of multiple users to obtain historical shopping behavior information of each of the users;

[0012] Based on a preset natural language model, each of the historical shopping behavior information is converted into a coding vector to obtain a user coding vector corresponding to each of the users, so as to construct the preset user vector database.

[0013] In a possible implementation, the historical shopping behavior information includes multiple historical shopping behavior descriptions; the user coding vector includes: a shopping behavior coding vector and a comprehensive preference coding vector;

[0014] The method of converting the historical shopping behavior information into a coding vector based on a preset natural language model to obtain a user coding vector corresponding to each user to construct the preset user vector database includes:

[0015] Based on a preset natural language model, vector encoding is performed on each of the historical shopping behavior descriptions to obtain a plurality of shopping behavior encoding vectors; the shopping behavior encoding vectors correspond one-to-one to the historical shopping behaviors;

[0016] Vector merging processing is performed on the plurality of shopping behavior coding vectors to obtain the comprehensive preference coding vector corresponding to each user individually, so as to construct the preset user vector database.

[0017] In a possible implementation, the performing similarity screening according to the target product coding vector and a preset user vector database to determine a plurality of groups of first user coding vectors matching the target product coding vector from the preset user vector database includes:

[0018] Calculating the vector similarity between the target product code vector and each of the user code vectors;

[0019] The user code vector whose vector similarity is higher than a preset first threshold is determined as the first user code vector.

[0020] In a possible implementation manner, determining the target user according to each of the first user coding vectors and the target product coding vector includes:

[0021] When there are multiple first user code vectors, determining a second user code vector according to a first vector similarity ranking; the first vector similarity ranking is generated based on the vector similarity of each of the first user code vectors;

[0022] Obtaining user activity corresponding to the second user encoding vector;

[0023] With respect to the user activity corresponding to the second user coding vector, a user whose user activity is greater than a preset second threshold is determined as the target user.

[0024] In a possible implementation, the target marketing scenarios include: new product promotion scenarios and old product clearance scenarios;

[0025] When the target marketing scenario is the new product promotion scenario, the target product coding vector includes: new product preference features; the new product preference features include: new product sensitivity, new product attention, and new product purchase tendency;

[0026] When the target marketing scenario is the old product clearance scenario, the target product coding vector includes: promotion activity preference features; the promotion activity preference features include: promotion sensitivity, promotion participation and promotion response rate.

[0027] In a second aspect, an embodiment of the present application provides a target user screening system, including:

[0028] A strategy determination module is used to perform screening strategy analysis based on target marketing scenarios and target product attribute information to determine target user screening strategies;

[0029] A first vector encoding module is used to perform vector encoding on the target user screening strategy based on a preset natural language model to obtain a target product encoding vector; the target product encoding vector is used to characterize the product features of the target product and the marketing preference of the target marketing scenario;

[0030] A first screening module is used to perform similarity screening based on the target product coding vector and a preset user vector database, so as to determine a plurality of groups of first user coding vectors matching the target product coding vector from the preset user vector database; the first user coding vector is used to characterize the historical shopping behavior characteristics of the user;

[0031] The second screening module is used to determine the target user according to each of the first user coding vectors and the target product coding vector.

[0032] In a possible implementation, the system further includes: a database construction module, wherein the database construction module is specifically configured to:

[0033] Performing data standardization processing on the historical shopping logs of multiple users to obtain historical shopping behavior information of each of the users;

[0034] Based on a preset natural language model, each of the historical shopping behavior information is converted into a coding vector to obtain a user coding vector corresponding to each of the users, so as to construct the preset user vector database.

[0035] In a possible implementation, the historical shopping behavior information includes multiple historical shopping behavior descriptions; the user coding vector includes: a shopping behavior coding vector and a comprehensive preference coding vector;

[0036] The database construction module is specifically used for:

[0037] Performing data standardization processing on the historical shopping logs of multiple users to obtain historical shopping behavior information of each of the users;

[0038] Based on a preset natural language model, each of the historical shopping behavior information is converted into a coding vector to obtain a user coding vector corresponding to each of the users, so as to construct the preset user vector database.

[0039] In a third aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any possible target user screening method in the first aspect.

[0040] Compared with the prior art, the present application has the following beneficial effects: the embodiment of the present application provides a method, system and medium for screening target users. In the method, it is first necessary to analyze the screening strategy according to the target marketing scenario and the target product attribute information, so as to combine the actual marketing needs with the attributes of the target product itself, thereby determining a target user screening strategy that is more suitable for the actual marketing scenario and improving the screening accuracy for the target user. Subsequently, the target product attribute information and the target user screening strategy are vector-encoded based on a preset natural language model. The natural language model can accurately capture the core semantics of the target product attributes and the target user screening strategy, and convert these core semantic features into a low-dimensional target product coding vector, so as to more finely capture the similarity relationship between the target user's historical shopping behavior and product features. Therefore, after determining the target product coding vector, it is screened for similarity with the preset user vector database, and multiple groups of user coding vectors matching the target product coding vector are determined, and the target user is screened and determined according to the similarity between each user coding vector and the target product coding vector. In this way, similarity is calculated through the encoding vector generated by the natural language model, and target users are screened based on the similarity between the two. The marketing characteristics of the product and the historical behavioral purchasing characteristics of the user can be intuitively converted into semantic data, and target users can be screened based on the similarity of the data at the semantic level. This can more comprehensively characterize the product characteristics of the target product and the shopping behavior preferences of different users, thereby improving the screening accuracy of target users. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0042] Figure 1A flowchart of a target user screening method provided in an embodiment of the present application;

[0043] Figure 2 A schematic diagram of a process for constructing a preset user vector database provided in an embodiment of the present application;

[0044] Figure 3 A flowchart of another target user screening method provided in an embodiment of the present application;

[0045] Figure 4 A schematic diagram of another target user screening method provided in an embodiment of the present application;

[0046] Figure 5 A schematic diagram of the structure of a target user screening system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of this application more clear, the following is a further detailed description of this application in combination with specific embodiments and with reference to the accompanying drawings. It should be noted that the embodiments described in the embodiments of this application are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0048] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should be understood by people with ordinary skills in the field to which the present application belongs. The "first", "second" and similar words used in the embodiments of the present application do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "include" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right", etc. are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0049] As described above, in the current scenario-based marketing plan, for different types of products, marketers need to screen potential target users who are willing to learn about the products based on their own experience. The accuracy of screening target users is completely determined by the marketers' experience, and the accuracy of screening target users is relatively poor.

[0050] In order to solve the above problems, the embodiment of the present application provides a method, system and medium for screening target users. In the method, it is first necessary to analyze the screening strategy according to the target marketing scenario and the target product attribute information, so as to combine the actual marketing needs with the attributes of the target product itself, thereby determining a target user screening strategy that is more suitable for the actual marketing scenario and improving the screening accuracy for the target user. Subsequently, the target product attribute information and the target user screening strategy are vector-encoded based on a preset natural language model. The natural language model can accurately capture the core semantics of the target product attributes and the target user screening strategy, and convert these core semantic features into a low-dimensional target product coding vector, so as to more finely capture the similarity relationship between the target user's historical shopping behavior and product features. Therefore, after determining the target product coding vector, it is screened for similarity with the preset user vector database, and multiple groups of user coding vectors that match the target product coding vector are determined, and the target user is screened and determined according to the similarity between each user coding vector and the target product coding vector. In this way, similarity is calculated through the encoding vector generated by the natural language model, and target users are screened based on the similarity between the two. The marketing characteristics of the product and the historical behavioral purchasing characteristics of the user can be intuitively converted into semantic data, and target users can be screened based on the similarity of the data at the semantic level. This can more comprehensively characterize the product characteristics of the target product and the shopping behavior preferences of different users, thereby improving the screening accuracy of target users.

[0051] In order to enable those skilled in the art to better understand the solution of the present application, 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 only 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.

[0052] 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 of relevant countries and regions.

[0053] Next, the target user screening method provided by the embodiment of the present application will be introduced in conjunction with the specific embodiment drawings, see Figure 1 , which is a flow chart of a target user screening method provided in an embodiment of the present application, and specifically includes the following steps:

[0054] S101: Perform screening strategy analysis based on target marketing scenarios and target product attribute information to determine target user screening strategies.

[0055] In actual product marketing scenarios, when formulating target user screening strategies, if only the attribute information of the target product (such as product name, product category, product feature description, etc.) is used, the target user screening strategies determined are often not in line with the actual product marketing needs of the marketer.

[0056] Therefore, in the preliminary screening stage of target users, the embodiment of the present application needs to analyze the screening strategy in combination with the target marketing scenario of the marketing party and the corresponding target product, so as to ensure the comprehensiveness of the target user screening strategy. Among them, the target marketing scenario can be roughly divided into a new product promotion scenario and an old product clearance scenario. In different marketing scenarios, the target product attribute information that needs to be extracted will also be different. For example, when the target marketing scenario is a new product promotion scenario, the target product attribute information will be closer to the time of listing of the target product, and the product difference between other products that are new during the same period. When the target marketing scenario is an old product clearance scenario, the target product attribute information will focus more on the promotion activities, promotional prices, and promotion duration of the target product. By classifying the target marketing scenario, the target product attribute information required for formulating the screening strategy can be extracted more accurately to ensure that the target user screening strategy can cover the key factors that may affect the user's purchase decision as comprehensively as possible, thereby realizing the comprehensive formulation of the target user screening strategy.

[0057] S102: Based on a preset natural language model, the target user screening strategy is vector-encoded to obtain a target product coding vector; the target product coding vector is used to characterize the product characteristics of the target product and the marketing preference of the target marketing scenario.

[0058] The natural language model can capture the deep semantic information in the text. Through the pre-trained preset natural language model, it can capture the subtle differences and complex relationships between the target product and the target marketing scenario, and convert the text version of the target user screening strategy into a high-dimensional (such as 768-dimensional) target product encoding vector, so as to more accurately characterize the specific characteristics of the product and the bias of the marketing strategy. For example, the text description of the attribute information of the target product may include multiple aspects such as the product's function, appearance, and price, while the description of the target marketing scenario may involve unstructured information such as the target user's interests and purchasing behavior. Through natural language processing technology, these complex and unstructured information can be effectively encoded into a target product encoding vector for the target product, thereby mathematically quantifying the product's characteristics and marketing preferences.

[0059] Specifically, in the process of vector encoding the target user screening strategy through the preset natural language model, in order to ensure the accuracy of the format of the language model input data, the target user screening strategy needs to be converted into a standardized description first. Take the above-mentioned new product promotion scenario and old product clearance scenario as an example:

[0060] New product promotion scenario: new product preference characteristics + target product attribute information + target user characteristics;

[0061] Clearance scenario for old products: promotion activity preference characteristics + target product attribute information + target user characteristics;

[0062] In this way, the standardized description processing of the target user screening strategy can ensure the format standardization of the input data for the preset natural language model. The standardized description of the target user screening strategy is input into the preset natural language model, and the corresponding encoding vector is generated using the encoding function in the language model. Thus, the complex marketing scenario requirements and target product attribute information are converted into a high-density target product encoding vector to improve the screening accuracy and efficiency of the target users.

[0063] In a possible implementation, the preset natural language model can use BERT, GPT, etc. as a basic language model. The embodiment of the present application further trains the language model by using various products and corresponding user information as training data to obtain a pre-trained preset natural language model.

[0064] S103: Perform similarity screening based on the target product coding vector and a preset user vector database to determine multiple groups of first user coding vectors matching the target product coding vector from the preset user vector database; the first user coding vector is used to characterize the user's historical shopping behavior characteristics.

[0065] The preset user vector database includes encoding vectors of multiple users. Among them, the encoding vector of the user is used as a characterization expression of the user's historical shopping behavior. For example, the user's preference for product types, the degree of preference for new products, and the degree of preference for promotional activities for old products, etc. The user encoding vector of each user in the preset user vector database can accurately characterize the shopping behavior preferences of each user, as well as the comprehensive preferences formed based on various types of shopping behavior preferences. Using this as the data benchmark for screening target users, the similarity screening link between the user encoding vector and the target product encoding vector can be refined, thereby accurately screening the first user encoding vector that has a certain similarity with the target product encoding vector.

[0066] Specifically, in the process of similarity screening based on the target product coding vector and the preset user vector database, it is necessary to calculate the vector similarity between the target product coding vector and each user coding vector, and screen the user coding vectors whose vector similarity is higher than the preset first threshold, and determine them as the first user coding vector.

[0067] Among them, the method for calculating the vector similarity between the target product coding vector and the user coding vector includes but is not limited to the cosine similarity calculation method, the Euclidean distance calculation method, the Manhattan distance calculation method and the hash similarity calculation method. In actual application scenarios, an adaptive similarity calculation method can be selected according to the computing resource situation and efficiency requirements, and this embodiment does not limit the method of similarity calculation.

[0068] In addition, in a possible implementation, in order to ensure the accuracy of the calculation of vector similarity, in addition to the target user screening strategy, the specific description of the target user can also be used as part of the reference data to generate the corresponding target product coding vector in the generation stage of the target product coding vector. The specific description of the target user includes the historical shopping behavior preferences and product attribute preferences specific to the target user. These feature descriptions specific to the target user are used as the basic data for generating the target product coding vector, so that when calculating the vector similarity between the target product coding vector and each user coding vector, more accurate and higher similarity comparison conditions can be obtained, thereby improving the calculation accuracy of the vector similarity, and then improving the accuracy of screening the target user.

[0069] From the description of the preset user vector database in step S103, it can be seen that the user coding vectors stored in the database for each user are the key to screening target users in the embodiment of the present application. In the user coding vector, the more accurate the description of the user's historical shopping behavior characteristics is, the higher the accuracy of screening the target user is. Therefore, the establishment of the preset user vector database is particularly important. Next, the construction process of the preset user vector database will be introduced in conjunction with the specific embodiment drawings.

[0070] See also Figure 2 , which is a schematic diagram of a process for constructing a preset user vector database provided in an embodiment of the present application, specifically comprising the following steps:

[0071] S1031: Perform data standardization processing on the historical shopping logs of multiple users to obtain historical shopping behavior information of each of the users.

[0072] The user's historical shopping log is used to characterize the user's historical shopping behavior and the relevant product information involved in the historical shopping behavior. The user's historical shopping behavior can be collected from various data sources, such as e-commerce platform databases, user behavior logs, etc. In the initial acquisition stage of the historical shopping log, it is necessary to perform data standardization on the historical shopping log to obtain historical shopping behavior information in a standard format. For example, the user's historical shopping behavior information can refer to the following format:

[0073] Product information|channel|cycle to market|promotion intensity|benefits form|;

[0074] In this way, through the user's historical shopping behavior information, we can characterize the shopping behavior preferences of different users and their own preferences for products. On this basis, these data are converted into a format suitable for natural language model analysis to facilitate the conversion of encoding vectors.

[0075] S1032: Based on a preset natural language model, convert each of the historical shopping behavior information into a coding vector to obtain a user coding vector corresponding to each of the users, so as to construct the preset user vector database.

[0076] In a possible implementation, the preset user vector database can be a relational database, a NoSQL database or a specialized vector database (such as Faiss, Milvus, etc.). The specific selection depends on the data scale, query speed requirements and system architecture design, which is not limited in this embodiment.

[0077] Furthermore, through the pre-trained preset natural language model, the historical shopping behavior information of each user is used as the input data of the preset natural language model, and the high-dimensional raw data is converted into a low-dimensional dense vector representation, thereby capturing the inherent characteristics of the user in the historical shopping behavior, so as to determine the user encoding vector for each user. The user encoding vector of each user is associated with the user's identity ID and stored in the user vector database, so as to realize the construction of the preset user vector database.

[0078] Specifically, in the process of building a preset user vector database, in order to ensure the data accuracy of each user coding vector in the database, the generation of the user coding vector needs to be obtained by performing vector coding analysis on each historical shopping behavior of the user. The process of building a preset user vector database in this way is implemented through the following two steps:

[0079] Step 1: Based on a preset natural language model, vector encoding is performed on each of the historical shopping behavior descriptions to obtain a plurality of shopping behavior encoding vectors; the shopping behavior encoding vectors correspond one-to-one to the historical shopping behaviors.

[0080] The user's historical shopping information includes multiple descriptions of the user's historical shopping behaviors. The user coding vector for the user includes a shopping behavior coding vector and a comprehensive preference coding vector. Among them, the shopping behavior coding vector is a coding vector corresponding to each historical shopping behavior description of the user. The vector coding for each historical behavior description of the user can accurately reflect the user's behavior preference represented by the historical shopping behavior under each specific historical shopping behavior, so as to more accurately build a user profile and determine the corresponding user coding vector.

[0081] Step 2: performing vector merging processing on the plurality of shopping behavior coding vectors to obtain the comprehensive preference coding vector corresponding to each user separately, so as to construct the preset user vector database.

[0082] On the other hand, the comprehensive coding vector is used to represent the comprehensive preferences of users. It is not limited to the user's historical shopping behavior preferences, but is more inclined to the user's preference for product information and various promotional activities. Vector merging of multiple shopping behavior coding vectors of users can more comprehensively represent the user's shopping preferences and shopping habits, thereby building a more accurate user portrait and improving the accuracy of target user screening.

[0083] It should be noted that in the user coding vector, the user's shopping behavior coding vector and the comprehensive preference coding vector together constitute the user coding vector, and the user coding vector does not only contain a single comprehensive preference coding vector. The significance of this setting is to improve the accuracy of similarity screening. When performing similarity screening, the user's shopping behavior coding vector and the comprehensive preference coding vector can be compared with the target product coding vector at the same time, and when at least one of the coding vectors has a certain similarity, it is summarized as the corresponding first user coding vector, so as to ensure the accuracy of similarity screening in step S103.

[0084] S104: Determine a target user according to each of the first user coding vectors and the target product coding vector.

[0085] When multiple first user code vectors with vector similarity greater than a preset first threshold are screened out, it is necessary to combine the similarity rankings of the multiple first user code vectors and the activity of each user to screen out the final target user, so as to further improve the screening accuracy of the target user. Next, the specific screening process of the target user will be introduced in conjunction with the specific embodiment drawings, see Figure 3 , which is a flow chart of another target user screening method provided by an embodiment of the present application, which specifically includes the following steps:

[0086] S1041: when there are multiple first user coding vectors, determine a second user coding vector according to a first vector similarity ranking; the first vector similarity ranking is generated based on the vector similarity of each of the first user coding vectors.

[0087] When there are multiple first user coding vectors whose vector similarity is higher than a preset first threshold, the multiple coding vectors are ranked according to their first vector similarities, and a coding vector ranked higher than a certain value is selected from the ranking and determined as the second user coding vector.

[0088] S1042: Obtain the user activity corresponding to the second user coding vector.

[0089] S1043: With respect to the user activity corresponding to the second user coding vector, determine a user whose user activity is greater than a preset second threshold as the target user.

[0090] In actual application scenarios, sometimes even if the user coding vector corresponding to the user has a high similarity with the target product coding vector, if the user has not been active for a long time, the product recommendation for the user is likely to fail to get a response. Therefore, on the basis of screening out the second user coding vector based on the first vector similarity ranking, it is also necessary to obtain the user activity corresponding to each second user coding vector. If there is a user whose user activity is lower than the preset second threshold, it means that the user has been dormant for a long time and is not determined as a target user. In addition, users whose user activity is greater than the preset second threshold among multiple second user coding vectors can be determined as target users, so as to achieve accurate screening of target users in specific marketing scenarios and target products.

[0091] As can be seen from the description of the target marketing scenario above, in the embodiment of the present application, the target marketing scenario can be divided into a new product promotion scenario and an old product clearance scenario. In different marketing scenarios, the corresponding marketing strategies will be different, and thus the vector features included in the target product coding vector will also be different.

[0092] According to the new product promotion scenario and the old product clearance scenario, the vector features in the target product coding vector can be divided into new product preference features and promotion activity preference features. Among them, the new product preference features include new product sensitivity, new product attention and new product purchase tendency. The following is a specific introduction to these three features:

[0093] New product sensitivity: users’ sensitivity to newly launched products, including browsing, favorites, adding to cart, and purchasing behaviors.

[0094] New product attention: The degree of user attention to new products can be measured by the time users stay on the new product page, the frequency of multiple browsing of new products, etc.

[0095] New product purchase tendency: The user's willingness to purchase a new product in the early stage of its launch can be reflected by the user's historical purchase history of new products.

[0096] On the other hand, the promotion preference characteristics include promotion sensitivity, promotion participation and promotion response rate. The following is a detailed introduction to these three characteristics:

[0097] Promotion sensitivity: the user's sensitivity to promotional activities, including reactions to coupons, discounts, and other promotional methods.

[0098] Promotion engagement: The frequency and depth of user engagement with promotions, such as the number of views, add-to-carts, and purchases made by users during the promotion period.

[0099] Promotion response rate: the user's response rate to promotional information, including the proportion of clicking on promotional links, participating in promotional activities and ultimately purchasing.

[0100] Thus, when the target product code and the user code vector are screened for similarity, the feature determined by the target marketing scenario is compared with the feature in the user code vector, thereby improving the efficiency of screening the target user. For example, when the target marketing scenario is a new product promotion scenario, the new product preference feature of each user is obtained from the preset user vector database, and the new product preference feature in the target product code vector is screened for similarity comparison, thereby improving the efficiency of screening the target user.

[0101] For further information, please refer to Figure 4 A schematic diagram of another target user screening method is shown in FIG. In the figure, embedding represents the corresponding generated encoding vector. By encoding the vectors of products and users and performing similarity screening based on the vector database, the corresponding target users can be accurately screened out from multiple users to achieve the effect of improving the accuracy of target user screening.

[0102] The embodiment of the present application provides a method for screening target users, in which the method first needs to perform screening strategy analysis based on the target marketing scenario and the target product attribute information, so as to combine the actual marketing needs with the attributes of the target product itself, thereby determining a target user screening strategy that is more in line with the actual marketing scenario and improving the screening accuracy for the target user. Subsequently, the target product attribute information and the target user screening strategy are vector-encoded based on a preset natural language model. The natural language model can accurately capture the core semantics of the target product attributes and the target user screening strategy, and convert these core semantic features into a low-dimensional target product coding vector, so as to more finely capture the similarity relationship between the target user's historical shopping behavior and product features. Therefore, after determining the target product coding vector, it is screened for similarity with the preset user vector database, and multiple groups of user coding vectors that match the target product coding vector are determined, and the target user is screened and determined based on the similarity between each user coding vector and the target product coding vector. In this way, similarity is calculated through the encoding vector generated by the natural language model, and target users are screened based on the similarity between the two. The marketing characteristics of the product and the historical behavioral purchasing characteristics of the user can be intuitively converted into semantic data, and target users can be screened based on the similarity of the data at the semantic level. This can more comprehensively characterize the product characteristics of the target product and the shopping behavior preferences of different users, thereby improving the screening accuracy of target users.

[0103] A target user screening system provided in an embodiment of the present application is introduced below. The target user screening system described below and the target user screening method described above can be referenced to each other.

[0104] See also Figure 5 , which is a schematic diagram of the structure of a target user screening system provided in an embodiment of the present application, specifically including the following modules:

[0105] A strategy determination module 100 is used to perform screening strategy analysis based on target marketing scenarios and target product attribute information to determine target user screening strategies;

[0106] A first vector encoding module 200 is used to perform vector encoding on the target user screening strategy based on a preset natural language model to obtain a target product encoding vector; the target product encoding vector is used to characterize the product features of the target product and the marketing preference of the target marketing scenario;

[0107] A first screening module 300 is used to perform similarity screening based on the target product coding vector and a preset user vector database, so as to determine a plurality of first user coding vectors matching the target product coding vector from the preset user vector database; the first user coding vector is used to characterize the historical shopping behavior characteristics of the user;

[0108] The second screening module 400 is used to determine the target user according to each of the first user coding vectors and the target product coding vector.

[0109] In a possible implementation, the system further includes: a database construction module, wherein the database construction module is specifically configured to:

[0110] Performing data standardization processing on the historical shopping logs of multiple users to obtain historical shopping behavior information of each of the users;

[0111] Based on a preset natural language model, each of the historical shopping behavior information is converted into a coding vector to obtain a user coding vector corresponding to each of the users, so as to construct the preset user vector database.

[0112] In a possible implementation, the historical shopping behavior information includes multiple historical shopping behavior descriptions; the user coding vector includes: a shopping behavior coding vector and a comprehensive preference coding vector;

[0113] The database construction module is specifically used for:

[0114] Performing data standardization processing on the historical shopping logs of multiple users to obtain historical shopping behavior information of each of the users;

[0115] Based on a preset natural language model, each of the historical shopping behavior information is converted into a coding vector to obtain a user coding vector corresponding to each of the users, so as to construct the preset user vector database.

[0116] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, an embodiment of the present application also provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the target user screening method described in any of the above embodiments.

[0117] The computer-readable media of the embodiments of the present application include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0118] The computer instructions stored in the storage medium of the above embodiment are used to enable the computer to execute the target user screening method for millimeter wave signals as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0119] It should be noted that each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for methods, systems and media, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments. The methods, systems and media described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components indicated as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0120] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for screening target users, characterized in that: include: Conduct screening strategy analysis based on target marketing scenarios and target product attribute information to determine target user screening strategies; Based on a preset natural language model, the target user screening strategy is vector-encoded to obtain a target product coding vector; the target product coding vector is used to characterize the product features of the target product and the marketing preference of the target marketing scenario; the target product coding vector includes a specific description of the target user, and the specific description is used to characterize the historical shopping behavior preference and product attribute preference of the target user; Performing similarity screening based on the target product coding vector and a preset user vector database to determine multiple groups of first user coding vectors matching the target product coding vector from the preset user vector database; the first user coding vector is used to characterize the historical shopping behavior characteristics of the user; Determining a target user according to each of the first user coding vectors and the target product coding vector; The method further comprises: Performing data standardization processing on the historical shopping logs of multiple users to obtain the historical shopping behavior information of each of the users; Based on a preset natural language model, each of the historical shopping behavior information is converted into a coding vector to obtain a user coding vector corresponding to each of the users to construct the preset user vector database; The determining the target user according to each of the first user coding vectors and the target product coding vector comprises: When there are multiple first user code vectors, determining a second user code vector according to a first vector similarity ranking; the first vector similarity ranking is generated based on the vector similarity of each of the first user code vectors; Obtaining user activity corresponding to the second user encoding vector; With respect to the user activity corresponding to the second user coding vector, a user whose user activity is greater than a preset second threshold is determined as the target user.

2. The method according to claim 1, characterized in that The historical shopping behavior information includes multiple historical shopping behavior descriptions; The user coding vector includes: a shopping behavior coding vector and a comprehensive preference coding vector; The method of converting the historical shopping behavior information into a coding vector based on a preset natural language model to obtain a user coding vector corresponding to each user to construct the preset user vector database includes: Based on a preset natural language model, vector encoding is performed on each of the historical shopping behavior descriptions to obtain a plurality of shopping behavior encoding vectors; the shopping behavior encoding vectors correspond one-to-one to the historical shopping behaviors; Vector merging processing is performed on the plurality of shopping behavior coding vectors to obtain the comprehensive preference coding vector corresponding to each user individually, so as to construct the preset user vector database.

3. The method according to claim 1, characterized in that The performing similarity screening according to the target product coding vector and a preset user vector database to determine a plurality of groups of first user coding vectors matching the target product coding vector from the preset user vector database includes: Calculating the vector similarity between the target product code vector and each of the user code vectors; The user code vector whose vector similarity is higher than a preset first threshold is determined as the first user code vector.

4. The method according to claim 1, characterized in that: The target marketing scenarios include: new product promotion scenarios and old product clearance scenarios; When the target marketing scenario is the new product promotion scenario, the target product coding vector includes: new product preference features; the new product preference features include: new product sensitivity, new product attention, and new product purchase tendency; When the target marketing scenario is the old product clearance scenario, the target product coding vector includes: promotion activity preference features; the promotion activity preference features include: promotion sensitivity, promotion participation and promotion response rate.

5. A target user screening system, characterized in that: include: A strategy determination module is used to perform screening strategy analysis based on target marketing scenarios and target product attribute information to determine target user screening strategies; A first vector encoding module is used to perform vector encoding on the target user screening strategy based on a preset natural language model to obtain a target product encoding vector; the target product encoding vector is used to characterize the product features of the target product and the marketing preference of the target marketing scenario; the target product encoding vector includes a specific description of the target user, and the specific description is used to characterize the historical shopping behavior preference and product attribute preference of the target user; A first screening module is used to perform similarity screening based on the target product coding vector and a preset user vector database, so as to determine a plurality of groups of first user coding vectors matching the target product coding vector from the preset user vector database; the first user coding vector is used to characterize the historical shopping behavior characteristics of the user; A second screening module, configured to determine a target user according to each of the first user coding vectors and the target product coding vector; The system further includes: a database construction module, wherein the database construction module is specifically used for: Performing data standardization processing on the historical shopping logs of multiple users to obtain the historical shopping behavior information of each of the users; Based on a preset natural language model, each of the historical shopping behavior information is converted into a coding vector to obtain a user coding vector corresponding to each of the users to construct the preset user vector database; The determining the target user according to each of the first user coding vectors and the target product coding vector comprises: When there are multiple first user code vectors, determining a second user code vector according to a first vector similarity ranking; the first vector similarity ranking is generated based on the vector similarity of each of the first user code vectors; Obtaining user activity corresponding to the second user encoding vector; With respect to the user activity corresponding to the second user coding vector, a user whose user activity is greater than a preset second threshold is determined as the target user.

6. The system according to claim 5, characterized in that The historical shopping behavior information includes multiple historical shopping behavior descriptions; the user coding vector includes: a shopping behavior coding vector and a comprehensive preference coding vector; The database construction module is specifically used for: Performing data standardization processing on the historical shopping logs of multiple users to obtain the historical shopping behavior information of each of the users; Based on a preset natural language model, each of the historical shopping behavior information is converted into a coding vector to obtain a user coding vector corresponding to each of the users, so as to construct the preset user vector database.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the target user screening method described in any one of claims 1 to 4 is implemented.

Images