Strategy generation method and device, computer equipment and readable storage medium

Abstract

The invention relates to a strategy generation method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a user product feature data set of a target user, and determining each target recommended product corresponding to the target user according to the user product feature data set and a product recommendation model; the user product feature data set comprises a user feature data set of a target user; according to the business personnel attribute data set, the verbal skill data set, the product attribute data set and the user feature data set, constructing a product verbal skill feature data set of each target recommended product; determining a target recommendation verbal skill corresponding to the target recommendation product based on the language model and the product verbal skill feature data set; and generating a product recommendation strategy for the target recommended product of the target user according to the target recommended product and the target recommended verbal skill corresponding to the target recommended product. By adopting the method, the accuracy of the product recommendation strategy can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a strategy generation method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] As people's insurance awareness gradually increases, insurance has become an important part of the current social security system. Business personnel will recommend insurance to target users based on product recommendation strategies, so that target users purchase insurance based on the recommendations of business personnel. Among them, the product recommendation strategy is generated in advance according to the strategy generation method.

[0003] In the current strategy generation method, since the product recommendation strategy includes the target recommended product and the target recommended speech, the product recommendation strategy generation process includes the target recommended product determination process and the target recommended speech generation process. Specifically, the business personnel determine the target recommended product corresponding to the target user in each product based on experience. Then, the business personnel select the target recommended speech corresponding to the target recommended product from several fixed speech, and generate the product recommendation strategy of the target recommended product based on the target recommended product and the target recommended speech.

[0004] However, the current strategy generation method has a large subjective factor in manually determining the target recommended products, resulting in low accuracy of the target recommended products. In addition, the manual determination of the target recommended words in a fixed wording is relatively limited. As a result, the accuracy of the product recommendation strategy is low. Summary of the invention

[0005] Based on this, it is necessary to provide a strategy generation method, apparatus, computer device, computer readable storage medium and computer program product to address the above technical issues.

[0006] In a first aspect, the present application provides a strategy generation method, comprising:

[0007] Acquire a user product feature data set of a target user, and determine each target recommended product corresponding to the target user according to the user product feature data set and a product recommendation model; the user product feature data set includes the user feature data set of the target user;

[0008] Constructing a product speech feature data set for each target recommended product based on the business personnel attribute data set, the speech data set, the product attribute data set and the user feature data set;

[0009] Determine the target recommendation speech corresponding to the target recommended product based on the language model and the product speech feature data set;

[0010] A product recommendation strategy for the target recommended product for the target user is generated according to the target recommended product and the target recommendation speech corresponding to the target recommended product.

[0011] In one embodiment, constructing a product speech feature dataset for each target recommended product based on the business personnel attribute dataset, the speech dataset, the product attribute dataset and the user feature dataset includes:

[0012] For each of the target recommended products, determine a second product attribute data subset of the target recommended product in the product attribute data set, and determine each to-be-recommended speech corresponding to the target recommended product in the speech data set;

[0013] Determining a second business personnel attribute data subset corresponding to the target user in the business personnel attribute data set;

[0014] A product speech feature data set of the target recommended product is constructed based on each of the to-be-recommended speech, the second product attribute data subset, the user feature data set and the second business personnel attribute data subset.

[0015] In one embodiment, obtaining a user product feature dataset of a target user includes:

[0016] Acquire a user feature data set of a target user; the user feature data set includes a user attribute data set, a historical purchase record data set, and an interest tag set of the target user; the historical purchase record data set includes a historical purchase record data subset of each historically purchased product;

[0017] For each of the historically purchased products, according to the first business person in the historical purchase record data subset of the historically purchased product, determining a first business person attribute data subset of the first business person in the business person attribute data set;

[0018] Determining a first product attribute data subset of the historically purchased product in the product attribute data set;

[0019] A user product feature data set of the target user is constructed based on the user feature data set, each of the first product attribute data subsets and each of the first business personnel attribute data subsets.

[0020] In one embodiment, obtaining a user feature dataset of a target user includes:

[0021] Obtain the user attribute dataset, historical purchase record dataset, and initial interest tag set of the target user;

[0022] Acquire a social media tag data set, and update the initial interest tag set according to the social media tag data set to obtain an interest tag set;

[0023] A user feature dataset of the target user is constructed according to the interest tag set, the user attribute dataset and the historical purchase record dataset.

[0024] In one embodiment, updating the initial interest tag set according to the social media tag data set to obtain the interest tag set includes:

[0025] Determine whether there is a target interest tag in the initial interest tag set that is the same as a target social media tag in the social media tag dataset;

[0026] If the target interest tag is the same as the target social media tag, determining an associated social media tag associated with the target social media tag in the social media tag dataset;

[0027] The associated social media tags are added to the initial interest tag set to obtain an interest tag set.

[0028] In one embodiment, determining each target recommended product corresponding to the target user according to the user product feature data set and the product recommendation model includes:

[0029] Based on the product recommendation model, predictive processing is performed on the user product feature data set and the product list to obtain a recommendation probability corresponding to each product in the product list;

[0030] Target recommended products are screened from among the products based on the recommendation probabilities of the products and preset recommendation conditions.

[0031] In one embodiment, generating a product recommendation strategy for the target recommended product for the target user based on the target recommended product and the target recommendation words corresponding to the target recommended product includes:

[0032] Obtaining the recommendation time of the target recommended product, and determining the target recommendation channel corresponding to the target recommended product in each channel;

[0033] A product recommendation strategy for the target recommended product of the target user is constructed based on the user feature data set of the target user, the target recommendation channel, the recommendation time, the second product attribute data subset corresponding to the target recommended product and the target recommendation words.

[0034] In a second aspect, the present application also provides a strategy generation device, including:

[0035] An acquisition module is used to acquire a user product feature data set of a target user, and determine each target recommended product corresponding to the target user according to the user product feature data set and a product recommendation model; the user product feature data set includes the user feature data set of the target user;

[0036] A construction module, used to construct a product speech feature data set for each target recommended product based on the business personnel attribute data set, the speech data set, the product attribute data set and the user feature data set;

[0037] A determination module, configured to determine a target recommendation speech corresponding to the target recommended product based on a language model and the product speech feature data set;

[0038] A generation module is used to generate a product recommendation strategy for the target recommended product for the target user based on the target recommended product and the target recommendation speech corresponding to the target recommended product.

[0039] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0040] Acquire a user product feature data set of a target user, and determine each target recommended product corresponding to the target user according to the user product feature data set and a product recommendation model; the user product feature data set includes the user feature data set of the target user;

[0041] Constructing a product speech feature data set for each target recommended product based on the business personnel attribute data set, the speech data set, the product attribute data set and the user feature data set;

[0042] Determine the target recommendation speech corresponding to the target recommended product based on the language model and the product speech feature data set;

[0043] A product recommendation strategy for the target recommended product for the target user is generated according to the target recommended product and the target recommendation speech corresponding to the target recommended product.

[0044] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0045] Acquire a user product feature data set of a target user, and determine each target recommended product corresponding to the target user according to the user product feature data set and a product recommendation model; the user product feature data set includes the user feature data set of the target user;

[0046] Constructing a product speech feature data set for each target recommended product based on the business personnel attribute data set, the speech data set, the product attribute data set and the user feature data set;

[0047] Determine the target recommendation speech corresponding to the target recommended product based on the language model and the product speech feature data set;

[0048] A product recommendation strategy for the target recommended product for the target user is generated according to the target recommended product and the target recommendation speech corresponding to the target recommended product.

[0049] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0050] Acquire a user product feature data set of a target user, and determine each target recommended product corresponding to the target user according to the user product feature data set and a product recommendation model; the user product feature data set includes the user feature data set of the target user;

[0051] Constructing a product speech feature data set for each target recommended product based on the business personnel attribute data set, the speech data set, the product attribute data set and the user feature data set;

[0052] Determine the target recommendation speech corresponding to the target recommended product based on the language model and the product speech feature data set;

[0053] A product recommendation strategy for the target recommended product for the target user is generated according to the target recommended product and the target recommendation speech corresponding to the target recommended product.

[0054] The above-mentioned strategy generation method, device, computer equipment, computer-readable storage medium and computer program product obtain a user product feature data set of a target user, and determine each target recommended product corresponding to the target user according to the user product feature data set and the product recommendation model; the user product feature data set includes the user feature data set of the target user; a product speech feature data set of each target recommended product is constructed according to the business personnel attribute data set, the speech data set, the product attribute data set and the user feature data set; the target recommendation speech corresponding to the target recommended product is determined based on the language model and the product speech feature data set; a product recommendation strategy for the target recommended product of the target user is generated according to the target recommended product and the target recommendation speech corresponding to the target recommended product. With this method, the target recommended product is determined by the product recommendation model and the user product feature data set that characterizes the association characteristics between the user and the product, the subjective factors in determining the target recommended product are reduced, and an accurate target recommended product is obtained. Then, by constructing a product speech feature data set, and determining the target recommendation speech corresponding to the target recommended product based on the language model and the objective product speech feature data set, human participation is avoided, the subjective factors in determining the target recommendation speech are reduced, and the accuracy of the target recommendation is improved. Therefore, generating a product recommendation strategy based on target recommendation words and target recommendation products improves the accuracy of the product recommendation strategy, thereby improving the accuracy of the strategy generation method. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions 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 related drawings can be obtained based on these drawings without paying creative work.

[0056] Figure 1 A schematic diagram of a process flow of a strategy generation method in an embodiment;

[0057] Figure 2 A schematic diagram of a process for constructing a product speech feature data set in one embodiment;

[0058] Figure 3 A schematic diagram of a process for obtaining a user product feature data set in one embodiment;

[0059] Figure 4 A schematic diagram of a process for obtaining a user feature data set in one embodiment;

[0060] Figure 5 A schematic diagram of a process for updating an initial interest tag set in one embodiment;

[0061] Figure 6 A schematic diagram of a process for determining a target recommended product in one embodiment;

[0062] Figure 7 A schematic diagram of a process for generating a product recommendation strategy in one embodiment;

[0063] Figure 8 A schematic diagram of the architecture of a policy generation system in one embodiment;

[0064] Fig. 9 is a structural block diagram of a strategy generating device in an embodiment;

[0065] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0067] In one embodiment, Figure 1 As shown, a policy generation method is provided. The embodiment of the present application takes the method applied to a computer device as an example for explanation. The embodiment of the present application does not limit the execution device for executing the policy generation method, and includes the following steps:

[0068] Step 102: Obtain a user product feature data set of the target user, and determine target recommended products corresponding to the target user based on the user product feature data set and the product recommendation model.

[0069] The user product feature data set includes the user feature data set of the target user. The user product feature data subset includes the user features, business personnel attribute features and product attribute features of the target user. User features include user attribute features, behavior features and intention features. The target user is the user to be recommended.

[0070] In implementation, the computer device obtains a business personnel attribute data set, a product attribute data set, and a speech data set from a database. The computer device obtains a user feature data set of the target user. The user feature data set contains the user attribute features, behavior features, and intention features of the target user. The computer device determines each first product attribute data subset and each first business personnel attribute data subset based on the user product feature data set, and constructs a user product feature data set corresponding to the target product based on the first product attribute data subset, the first business personnel attribute data subset, and the user feature data set. The computer device processes the user product feature data set and the product list based on the user product recommendation model to determine each target recommended product corresponding to the target user.

[0071] Specifically, the computer device traverses the user list according to a preset time period, determines each user in the user list as a target user, and obtains a user feature data set of the target user. Then, the computer device determines each first product attribute data subset associated with the user feature data set in the product attribute data set, and determines each first business personnel attribute data subset associated with the user feature data set in the business personnel attribute data set. The computer device constructs a user product feature data set for the target user based on each first product attribute data subset, each first business personnel attribute data subset, and the user feature data set. Then, the computer device processes the user product feature data subset and the product list to obtain the recommendation probability of each product in the product list, and determines the target recommended product corresponding to the target user based on the recommendation probability of each product.

[0072] Step 104, constructing a product speech feature dataset for each target recommended product based on the business personnel attribute dataset, the speech dataset, the product attribute dataset and the user feature dataset.

[0073] Among them, the user feature data set includes a historical purchase record data set. The speech data set includes each speech and the corresponding relationship between each product and each speech. The second business personnel attribute data subset includes the business personnel attribute characteristics of the second business personnel. The user feature data set includes the user attribute characteristics, behavior characteristics and intention characteristics of the target user. The second product attribute data subset includes the product attribute characteristics of the target recommended product.

[0074] In implementation, the computer device determines the second business personnel attribute data subset corresponding to the target user in the business attribute data set based on the historical purchase record data set. At the same time, for each target recommended product, the computer device determines the second product attribute data subset corresponding to the target recommended product in the product attribute data set, and determines each to-be-recommended speech corresponding to the target recommended product in the speech data set. Then, the computer device constructs a product speech feature data set for the target recommended product based on each to-be-recommended speech, the second business personnel attribute data subset, the user feature data set, and the second product attribute data subset. Since the product speech feature data set contains the second business personnel attribute data subset, the user feature data set, and the second product attribute data subset. Therefore, the product speech feature data set contains the business personnel attribute characteristics of the second business personnel, the user attribute characteristics, behavior characteristics, and intention characteristics of the target user, and the product attribute characteristics of the target recommended product.

[0075] Step 106: determine the target recommendation words corresponding to the target recommendation product based on the language model and the product word feature data set.

[0076] The product speech feature dataset contains all the speech to be recommended. The language model is obtained by training the NLP (Neuro-Linguistic Programming, Natural Language Processing) model based on the product speech sample dataset.

[0077] During implementation, the computer device inputs the product speech feature data set into the language model, processes the second business personnel attribute data subset, the user feature data set, and the second product attribute data subset in the product speech feature data set through the language model, and determines the target recommended speech corresponding to the target recommended product in each speech to be recommended through the language model.

[0078] Specifically, the computer device obtains a product speech sample data set. The product speech sample data set contains each product speech sample data subset. Each product speech sample data subset is a data subset generated when a sample data recommends a sample recommended product. The product speech sample data subset contains a sample user feature data set of a sample user, a sample product attribute data subset of a sample recommended product, a sample business staff attribute data subset of a sample business staff who recommends a sample product, and a sample recommended speech when recommending this recommended sample product. The computer device trains an NLP model based on the product speech sample data set, so that the NLP model can learn the association between user features (including user attribute features, behavior features, and intention features), business staff features, product attribute features, and speech features, and obtain a language model. The computer device inputs the product speech feature data set into the language model, and performs prediction processing on the product speech feature set through the language model, so that the language model can determine the target recommended speech corresponding to the target recommended product according to the business staff attribute features of the second business staff in the product speech feature set, the user attribute features, behavior features, and intention features of the target user, and the product attribute features of the target recommended product.

[0079] In an exemplary embodiment, the computer device obtains a data subset generated by each sample user when recommending each recommended sample product, and obtains a product speech sample data subset. The recommended sample product is a historically purchased product of the sample user. Since the product speech sample data subset includes a sample user feature data set of the sample user, a sample product attribute data subset of the sample recommended product, a sample business staff attribute data subset of the sample business staff who recommended the sample product, and a sample recommended speech when recommending this recommended sample product, the product speech sample data subset includes user features of the sample user (including user attribute features, behavior features, and intention features of the sample user), business staff attribute features of the sample business staff, product attribute features, and speech features. The computer device determines the product speech sample data set as a speech training sample set, and trains the NLP model based on the speech training sample set to obtain a language model. During the training process of the NLP model, the computer device can use the GPU (Graphics Processing Unit) to accelerate parallel computing and improve training efficiency. At the same time, the computer device optimizes the model performance based on hyperparameter search (such as Grid Search or Bayesian Optimization) and verifies the generalization ability of the model through cross-validation technology to prevent overfitting. The computer device continuously monitors the training effect of the NLP model and judges the quality of the NLP model through model performance evaluation indicators (such as accuracy and recall rate). When the model performance evaluation indicators reach the preset deep learning model training stop condition, the computer device stops training the NLP model and determines the NLP model as a language model.

[0080] Optionally, the NLP model may be, but is not limited to, a causal language model (CLM) or a masked language model (MLM). The embodiments of the present application do not limit the NLP model.

[0081] Step 108, generating a product recommendation strategy for the target recommended product for the target user based on the target recommended product and the target recommendation script corresponding to the target recommended product.

[0082] In implementation, the computer device obtains a recommendation strategy template, and obtains the recommendation time and target recommendation channel according to the recommendation strategy template. Then, the computer device updates the recommendation strategy template based on the recommendation time, target recommendation channel, second recommended product, and target recommendation words corresponding to the target recommended product, and obtains the product recommendation strategy for the target recommended product of the target user. Then, the computer device pushes the product recommendation strategy of the target recommended product to the second business personnel according to the recommendation time and target recommendation channel, so that the second business personnel recommends the target recommended product to the target user according to the product recommendation strategy of the target recommended product.

[0083] In the above strategy generation method, the target recommended product is determined by the product recommendation model and the user-product feature data set that characterizes the association characteristics between users and products, which reduces the subjective factors in determining the target recommended product and obtains accurate target recommended products. Then, by constructing a product speech feature data set and determining the target recommendation speech corresponding to the target recommended product based on the language model and the objective product speech feature data set, human participation is avoided, the subjective factors in determining the target recommendation speech are reduced, and the accuracy of the target recommendation is improved. Therefore, the product recommendation strategy is generated based on the target recommendation speech and the target recommended product, which improves the accuracy of the product recommendation strategy and thus improves the accuracy of the strategy generation method.

[0084] In an exemplary embodiment, Figure 2 As shown, the specific processing process of step 104 includes steps 202 to 206. Among them:

[0085] Step 202: for each target recommended product, determine a second product attribute data subset of the target recommended product in the product attribute data set, and determine each to-be-recommended speech corresponding to the target recommended product in the speech data set.

[0086] The speech data set includes various speech techniques and the corresponding relationship between each speech technique and each product.

[0087] In implementation, for each target recommended product, the computer device searches the product attribute data set for the second product attribute data set corresponding to the target recommended product according to the product identifier of the target recommended product. At the same time, the computer device searches the speech data set for each speech to be recommended corresponding to the target recommended product according to the product identifier of the target recommended product.

[0088] Specifically, the target recommended product is a product that needs to be recommended to the target user. For each target recommended product, the computer device queries the product attribute data set for a second product attribute data subset corresponding to the target recommended product according to the product identifier of the target recommended product. The second product attribute data subset includes the product identifier, product type, product resources, applicable objects, and product introduction of the target recommended product. At the same time, the computer device queries the speech data set for each speech to be recommended corresponding to the target recommended product according to the product identifier of the target product. Among them, the speech in the speech data set is automatically generated and stored according to the communication records of the business personnel.

[0089] In an exemplary embodiment, the target recommended product is target recommended insurance. When the business personnel successfully recommends insurance, a visit list will be registered to obtain visit list data. The computer device extracts the salesperson's words for recommending insurance from the visit list data and adds the words to the words data set. For each target recommended insurance, the computer device queries the insurance attribute database for a second insurance attribute data subset of the target recommended insurance according to the insurance identifier of the target recommended insurance. The second insurance attribute data subset contains insurance identifiers, insurance types, insurance resources, applicable objects, and insurance introductions. Among them, insurance types include accident insurance, health insurance, and car insurance. Applicable objects include, for example, families, individuals, and car owners. At the same time, the computer device queries the words to be recommended corresponding to the target recommended insurance in the words data set according to the insurance identifier of each target recommended insurance. Each word to be recommended is an alternative word for recommending the target recommended insurance.

[0090] In an optional embodiment, the business personnel recommends insurance to the user by means of terminal voice. The terminal records the voice conversation between the business personnel and the user and converts the voice conversation into a text record. When the business personnel successfully recommends the insurance to the user, the terminal sends the text record to the computer device. The computer device receives the text record and determines that the text record is a speech. Then, the computer device adds the speech to the speech data set in the database.

[0091] Step 204: determine a second business personnel attribute data subset corresponding to the target user in the business personnel attribute data set.

[0092] The user feature data set includes a historical purchase record data set of the target user. The historical purchase record data set includes a historical purchase record data subset of each historically purchased product. The historical purchase record data subset includes the business person identification of the business person who recommended the historically purchased product.

[0093] In implementation, the computer device determines the business person identification with the most occurrences in the historical purchase record data set. Then, the computer device determines the business person identification as the second business person identification, and determines the business person corresponding to the business person identification as the second business person corresponding to the target user. Then, the computer device queries the business person attribute data set for the second business person attribute data subset based on the second business person identification.

[0094] In an exemplary embodiment, the historical purchase record data set includes a subset of historical purchase record data of the target user purchasing each insurance product. The computer device determines that the business personnel identifier with the most occurrences in each historical purchase record data set is the second business personnel identifier, and determines that the business personnel represented by the second business personnel identifier is the second business personnel. The second business personnel is the business personnel most familiar to the target user and is also the business personnel responsible for the target user. Based on the second business personnel identifier, the computer device queries the second business personnel attribute data subset in the business personnel attribute data set, thereby obtaining the business personnel attribute characteristics of the second business personnel. Among them, the second business personnel attribute data subset includes the organizational structure, business transaction volume, gender, age, languages ​​mastered, and place of origin of the second business personnel.

[0095] In an optional embodiment, the correspondence between each user and each business person is pre-set in the computer device. One business person is responsible for multiple users. The computer device determines the second business person corresponding to the target user according to the correspondence between each user and each business person. Then, the computer device determines the second business person attribute data subset of the second business person in the business person attribute data set.

[0096] Step 206, constructing a product speech feature data set of the target recommended product based on each speech to be recommended, the second product attribute data subset, the user feature data set and the second business personnel attribute data subset.

[0097] In implementation, the computer device combines the to-be-recommended speech, the second product attribute data subset, the user feature data set, and the second business personnel attribute data subset into a product speech feature data subset for each to-be-recommended speech. Then, the computer device combines each product speech feature data subset into a product speech feature data set for the target recommended product.

[0098] In this embodiment, by determining each to-be-recommended speech, each candidate speech is obtained. And by determining the second business personnel attribute data subset, the second product attribute data subset, each to-be-recommended speech and the user feature data set, a product speech feature data set is constructed, and the business personnel attribute characteristics of the second business personnel, the user attribute characteristics, behavior characteristics and intention characteristics of the target user and the product attribute characteristics of the target recommended product are obtained, which facilitates the subsequent determination of the target recommended speech based on the objective product speech feature data set, and improves the accuracy of the target recommended speech.

[0099] In an exemplary embodiment, Figure 3 As shown, the specific processing process of obtaining the user product feature data set of the target user in step 102 includes steps 302 to 308. Among them:

[0100] Step 302: Obtain a user feature data set of the target user.

[0101] The user feature data set includes a user attribute data set, a historical purchase record data set, and an interest tag set of the target user; the historical purchase record data set includes a subset of historical purchase record data for each historically purchased product.

[0102] In implementation, when it is necessary to recommend products to a target user, the computer device obtains a user attribute dataset, a historical purchase dataset, and an initial interest tag set of the target user. At the same time, the computer device obtains a social media tag dataset. The computer device updates the initial interest tag set according to the social media tag dataset to obtain an interest tag set. Then, the computer device combines the user attribute dataset, the historical purchase record dataset, and the interest tag set into a user feature dataset of the target user.

[0103] Specifically, the computer device obtains the user attribute data set, the historical purchase record data set and the initial interest tag set of the target user from the database according to the user identifier of the target user. The user attribute data set represents the user attribute characteristics of the target user. The user attribute data set includes the user identifier, user name, user gender, user age, user marital status, user car purchase status, etc. The historical purchase record data set includes a historical purchase record data subset of each historically purchased product, and the historical purchase record data subset includes the product identifier, recommended business personnel, product type, transaction status, etc. of the historically purchased product. The historical purchase record data set represents the behavioral characteristics of the target user. The initial interest tag set includes the browsing tags and behavioral tags of the target user. The initial interest tag set represents the behavioral characteristics and intentional characteristics of the target user. The computer device obtains the social media interest tag set, and updates the initial interest tag set according to the social media interest tag set to obtain the interest tag set. The computer device constructs the user feature data set of the target user based on the user attribute data set, the historical purchase record data set and the interest tag set.

[0104] Step 304 , for each historically purchased product, according to the first business person in the historical purchase record data subset of the historically purchased product, determine a first business person attribute data subset of the first business person in the business person attribute data set.

[0105] In implementation, the computer device determines, for each historically purchased product, a first business person in the historical purchase record data subset of the historically purchased product. The first business person is a business person who recommends the historically purchased product to the target user. The computer device queries the first business person attribute data subset of the first business person in the business person attribute data set, thereby obtaining the business person attribute characteristics of the business person corresponding to the historically purchased product and the target user.

[0106] In an exemplary embodiment, the historically purchased product is historically purchased insurance. For each historically purchased insurance, the computer device determines a first business person who recommends the historically purchased insurance in the historical purchase record data subset of the historically purchased insurance. Then, the computer device determines a first business person attribute data subset corresponding to the first business person in the business person attribute data set.

[0107] Step 306: determine a first product attribute data subset of historically purchased products in the product attribute data set.

[0108] In implementation, the computer device determines a first product attribute data subset of the historically purchased product in the product attribute data set according to the product identifier of the historically purchased product. The first product attribute data subset is a product attribute feature of the historically purchased product.

[0109] In an exemplary embodiment, the product is insurance. The computer device searches for a first insurance attribute data subset of the historically purchased insurance in the insurance attribute data set according to the insurance identification of the historically purchased insurance, thereby clarifying the product attribute characteristics of the historically purchased insurance.

[0110] Step 308: construct a user product feature data set of the target user based on the user feature data set, each first product attribute data subset and each first business personnel attribute data subset.

[0111] In implementation, the computer device constructs the product attribute data subset, the first business personnel attribute data subset corresponding to the product attribute data subset, and the user feature data set into a user product feature data subset for each first product attribute data subset. By constructing the user product feature data subset, the user attribute characteristics of the target user, the product characteristics of the historically purchased products, and the business attribute characteristics of the first business personnel who recommends the historically purchased products can be associated together. The computer device combines the user product feature attribute data subsets together to obtain a user product feature data set for the target user. The user product feature data set for the target user contains user attribute characteristics, product attribute characteristics, and business personnel attribute characteristics, as well as the association relationship between the user attribute characteristics, product attribute characteristics, and business personnel attribute characteristics.

[0112] In an exemplary embodiment, the target user is user a. The computer device obtains a user feature data set of user a. The user feature data set of user a includes a user attribute data set, a historical purchase record data set, and an initial interest tag set of user a. The historical purchase record data set of user a includes a historical purchase product data subset of the user's historical purchase of product X, the user's historical purchase of product Y, and the user's historical purchase of product Z. The computer device determines a first business person attribute data subset of the first business person in the business person attribute data set based on the first business person who recommended the historical purchase of product X in the historical purchase record data subset of the historical purchase of product X. Then, the computer device determines a first product attribute data subset of the historical purchase of product X in the product attribute data set. The computer device constructs the product attribute data subset of the historical purchase of product X, the first business person attribute data subset corresponding to the product attribute data subset, and the user feature data set as a user product feature data subset of the historical purchase of product X.

[0113] The computer device determines a first business personnel attribute data subset of the first business personnel in the business personnel attribute data set according to the first business personnel who is recommended to have historically purchased product Y in the historical purchase record data subset of the historical purchase product Y. Then, the computer device determines a first product attribute data subset of the historical purchase product Y in the product attribute data set. The computer device constructs the product attribute data subset of the historical purchase product Y, the first business personnel attribute data corresponding to the product attribute data subset, and the user feature data set into a user product feature data subset of the historical purchase product Y.

[0114] The computer device determines a first business personnel attribute data subset of the first business personnel in the business personnel attribute data set according to the first business personnel who recommended the historical purchase of product Z in the historical purchase record data subset of the historical purchase of product Z. Then, the computer device determines a first product attribute data subset of the historical purchase of product Z in the product attribute data set. The computer device constructs the product attribute data subset of the historical purchase of product Z, the first business personnel attribute data subset corresponding to the product attribute data subset, and the user feature data set into a user product feature data subset of the historical purchase of product Z.

[0115] The computer device constructs a product feature data subset of historically purchased product X, a product feature data subset of historically purchased product Y, and a product feature data subset of historically purchased product Z into a user product feature data set of user a.

[0116] In this embodiment, a user product feature data set is constructed by using data of different types and sources including a user feature data set, each first product attribute data subset, and each first business personnel attribute data subset, thereby enriching the user product feature data set, facilitating subsequent determination of target recommended products based on the objective user product feature data set, and improving the accuracy of target recommended products.

[0117] In an exemplary embodiment, Figure 4 As shown, the specific processing process of step 302 includes steps 402 to 406. Among them:

[0118] Step 402: Obtain a user attribute data set, a historical purchase record data set, and an initial interest tag set of the target user.

[0119] In implementation, the computer device obtains the user attribute data set and the historical purchase record data set of the target user from the database according to the user identification of the target user. At the same time, the computer device obtains the initial interest tag set of the target user.

[0120] Specifically, during the transfer of product resources, the business personnel will maintain the customer management system and manage the business personnel attribute data set and the user attribute data set in the database by maintaining the customer management system. The computer device obtains the user attribute data set and the business personnel attribute data set from the database through the customer management system. At the same time, the computer device obtains the historical purchase record data set of the target user from the enterprise's database according to the user ID of the target user. The computer device obtains the initial interest tag set of the target user from the third-party data platform. The initial interest tag set contains various interest tags of the target user, and the interest tags are divided into the behavior tags of the target user and the browsing tags of the target user.

[0121] In an optional embodiment, a third-party data platform, a customer management system, a database, and social media are data sources. The computer device establishes connections with multiple data sources through API (Application Programming Interface) and ETL (Extract, Transform, Load) tools (such as Apache NiFi, a powerful data flow management tool) to ensure the real-time and accuracy of data collection. The computer device obtains the user attribute data set, historical purchase record data set, and initial interest tag set of the target user from the data source. The computer device performs data cleaning and data standardization on the user attribute data set, historical purchase record data set, and initial interest tag set based on the data cleaning tool to obtain the processed user attribute data set, the processed historical purchase record data set, and the processed initial interest tag set. Among them, data cleaning includes but is not limited to removing duplicate records, filling missing values, data standardization, and data formatting. Through data formatting, the above data can be converted into a unified format to facilitate subsequent data processing. The computer device stores the processed user attribute data set, historical purchase record data set, and initial interest tag set in a distributed database to support efficient storage and retrieval of large-scale data.

[0122] Optionally, the third-party data platform may be, but is not limited to, a shopping platform, and the data cleaning tools include, but are not limited to, Pandas (a data analysis library) and Apache Spark (a big data processing engine). Distributed databases include, but are not limited to, Apache HDFS (a distributed file system) or Cassandra (a highly scalable, highly available distributed database). The present application embodiment does not limit the third-party data platform, distributed database, and data cleaning tool.

[0123] Step 404: obtain a social media tag data set, and update the initial interest tag set according to the social media tag data set to obtain an interest tag set.

[0124] In implementation, a computer device obtains an initial social media tag data set from a social media platform. The initial social media tag data set contains initial social media tags for each object. The initial social media tags are determined based on the user's behavior using the social media platform. The computer device performs clustering processing on the initial social media tag data set to obtain clusters. Each cluster contains an object group, and the object group contains initial social media tags for multiple objects. For each object group, the computer device extracts each social media tag in the object group according to a preset extraction threshold, and establishes an association relationship between each social media tag. Then, the computer device constructs a social media tag data set based on each associated social media tag. The computer device determines a target social media tag in the social media tag set, and updates the initial interest tag set based on the social media tag associated with the target social media tag to obtain an interest tag set.

[0125] Specifically, the computer device obtains an initial media tag data set from the social media through a social media API. Then, the computer device performs clustering processing on the initial social media tag data set according to a clustering algorithm to obtain each cluster cluster. Each cluster cluster contains an object group, and the object group contains objects of the same type and initial social media tags of each object. For each cluster cluster, the computer device counts the number of occurrences of each initial social media tag in the cluster cluster and each initial social media tag. Then, the computer device sorts each initial social media tag in order of the number of occurrences from large to small to obtain an initial social media tag sequence. Then, the computer device determines each social media tag in the initial social media tag sequence according to a preset extraction threshold, and establishes an association relationship between each social media tag. Then, the computer device constructs a social media tag set according to each social media tag under each cluster cluster. Then, the computer device determines a target social media tag in the social media tag set according to the target interest tag in the initial interest tag set, and updates the initial interest tag set according to the social media tag associated with the target social media tag to obtain an interest tag set.

[0126] Optionally, the extraction threshold may be, but is not limited to, set to 3. The extraction threshold is determined according to the policy generation requirement, and the embodiment of the present application does not limit the extraction threshold.

[0127] Optionally, the clustering algorithm may be, but is not limited to, a K-means algorithm (K-means clustering algorithm) or a FCM algorithm (Fuzzy C-means clustering algorithm). The number of center points of the clustering algorithm is determined according to the clustering requirements. The larger the number of center points of the clustering algorithm is set, the more clusters there are, the more social media tags are extracted, and the larger the social media tag data set is. The embodiments of the present application do not limit the clustering algorithm and the number of center points.

[0128] Step 406: construct a user feature dataset of the target user based on the interest tag set, the user attribute dataset, and the historical purchase record dataset.

[0129] In implementation, the computer device combines the interest tag set, the user attribute data set and the historical purchase record data set to obtain the user feature data set of the target user. By combining the interest tag set, the user attribute data set and the historical purchase record data set, the user attribute characteristics, behavior characteristics and intention characteristics of the target user can be associated.

[0130] In this embodiment, interest tag sets, user attribute data sets, and historical purchase record data sets are obtained from different data sources to achieve high-quality input of multi-source heterogeneous data, and a user feature data set of the target user is constructed based on the interest tag sets, user attribute data sets, and historical purchase record data sets, thereby enriching user features and improving the accuracy of the user feature data set.

[0131] In an exemplary embodiment, Figure 5 As shown, the specific process of updating the initial interest tag set according to the social media tag data set in step 404 to obtain the interest tag set includes steps 502 to 506. Among them:

[0132] Step 502 , determining whether there is a target interest tag in the initial interest tag set that is the same as a target social media tag in the social media tag dataset.

[0133] In implementation, the computer device determines, for each interest tag in the initial interest tag set, whether there is a social media tag in the social media tag data set that is the same as the interest tag. If there is a social media tag that is the same as the interest tag, the computer device determines the social media tag as the target social media tag and determines the interest tag as the target interest tag.

[0134] In an optional embodiment, if there is no target interest tag in the initial interest tag set that is identical to the target social media tag in the social media tag data, the computer device determines the initial interest tag set as the interest tag set and continues to execute the above step 406.

[0135] Step 504: if there is a target interest tag that is the same as the target social media tag, determine the associated social media tag associated with the target social media tag in the social media tag dataset.

[0136] In an implementation, if a target interest tag in the initial interest tag set is the same as a target social media tag in the social media tag set, the computer device determines an associated social media tag associated with the target social media tag in the social media tag set.

[0137] Step 506 , adding the associated social media tags to the initial interest tag set to obtain an interest tag set.

[0138] In implementation, the computer device adds the associated social media tag to the initial interest tag set to obtain the interest tag set. If there are multiple target social media tags, the computer device determines, for each target social media tag, the associated social media tag associated with the target social media tag in the social media tag set, and adds the associated social media tag to the initial interest tag set to obtain the interest tag set.

[0139] In an exemplary embodiment, the interest tags included in the initial interest tag set are vehicle, hospital, and beauty. The social media tag set includes vehicle and vehicle-related auto insurance and worry-free driving insurance, hospital and basic medical insurance and supplementary medical insurance associated with the hospital. The computer device determines that the vehicle exists in the social media tag data set for the interest tag of vehicle, and determines the vehicle as the target interest tag. The computer device determines the auto insurance and worry-free driving insurance associated with the vehicle in the social media tag data set, and adds the auto insurance and worry-free driving insurance to the initial interest tag set. The computer device determines that the hospital exists in the social media tag data set for the interest tag of hospital, and determines the hospital as the target interest tag. The computer device determines the basic medical insurance and supplementary medical insurance associated with the hospital in the social media tag data set, and adds the basic medical insurance and supplementary medical insurance to the initial interest tag set. The computer device determines that beauty does not exist in the social media tag set for the interest tag of beauty. The computer device determines the updated initial interest tag set as the interest tag set. The interest tag set includes vehicle, auto insurance, worry-free driving insurance, basic medical insurance, supplementary medical insurance, hospital, and beauty.

[0140] In this embodiment, the initial interest tag set is updated through the social media tag set to obtain the interest tag set, which enriches the interest tags in the interest tag set and improves the accuracy of the interest tag set.

[0141] In an exemplary embodiment, Figure 6As shown, the specific processing process of determining each target recommended product corresponding to the target user according to the user product feature data set and the product recommendation model in step 102 includes steps 602 to 604. Among them:

[0142] Step 602: Perform prediction processing on the user product feature data set and the product list based on the product recommendation model to obtain the recommendation probability corresponding to each product in the product list.

[0143] Among them, the product recommendation model is obtained by training a deep learning model with a Transformer (a deep learning model architecture) architecture based on a user product feature sample dataset.

[0144] In implementation, the computer device inputs the user product feature data set and the product list into the product recommendation model, and processes the user product feature data set through the product recommendation model to obtain the recommendation probability corresponding to each product in the product list. The recommendation probability corresponding to each product represents the conversion probability of the second business person recommending the product to the target user.

[0145] Specifically, the computer device obtains a user product feature sample data set, and trains a deep learning model based on the user product feature sample data set, so that the deep learning model can learn the association between user attribute features, behavior features, intention features, business personnel attribute features and product attribute features, and obtain a product recommendation model. The computer device inputs the user product feature data set and the user list into the product recommendation model, so that the product recommendation model predicts the recommendation probability of each product in the product list based on the user attribute features, behavior features and intention features of the target user in the user product feature data set, the second business personnel attribute features, and the product attribute features.

[0146] In an exemplary embodiment, a large-scale pre-trained model based on TensorFlow is pre-set in the computer device. The computer device determines the large-scale pre-trained model based on TensorFlow as a deep learning model. It should be noted here that the deep learning model needs to implement multi-task learning, that is, to simultaneously process tasks such as customer portrait construction, sentiment analysis, and intent recognition in one model. The computer device obtains a user product feature sample data set. The user product feature sample data set contains a user product feature sample data subset of each sample user. The process of constructing the user product feature sample data subset is the same as the process of the user product feature subset. For the specific process, please refer to the description of steps 402 to 406 above, which will not be repeated in the embodiments of the present application. The computer device determines the user product feature sample feature set as a product training sample set, and trains a deep learning model based on the product training sample set to obtain a product recommendation model. During the training process of deep learning models, computer equipment can use GPU to accelerate parallel computing and improve training efficiency. At the same time, computer equipment optimizes model performance based on hyperparameter search (such as grid search or Bayesian optimization) and verifies the generalization ability of the model through cross-validation technology to prevent overfitting. Computer equipment continuously monitors the training effect of deep learning models and judges the quality of deep learning models through model performance evaluation indicators (such as accuracy and recall rate). When the model performance evaluation indicators reach the preset deep learning model training stop conditions, the computer equipment stops training the deep learning model and determines the deep learning model as the product recommendation model.

[0147] Optionally, the deep learning model can be, but is not limited to, a large-scale pre-trained model built based on TensorFlow or PyTorch. The embodiments of the present application do not limit the deep learning model.

[0148] Step 604: Filter target recommended products from among the products based on the recommendation probability of each product and the preset recommendation conditions.

[0149] The recommendation condition is that the recommendation probability of the target recommended product is greater than or equal to the recommendation probability threshold and the sequence position of the target recommended product in the product sequence is greater than the position threshold.

[0150] In implementation, the computer device is pre-set with recommendation conditions. The computer device sorts the products in descending order of recommendation probability to obtain a product sequence. The computer device determines each initial target recommended product in the product sequence according to a preset position threshold. Then, for each initial target recommended product, the computer device determines whether the recommendation probability of the initial target recommended product is greater than or equal to a preset recommendation probability threshold. If the recommendation probability of the initial target recommended product is greater than or equal to the recommendation probability threshold, the computer device determines the initial target recommended product as the target recommended product.

[0151] Optionally, the recommendation probability threshold can be but is not limited to being set to 90%, and the position threshold can be but is not limited to being set to the 10th place. The recommendation probability threshold and the position threshold are set according to the recommendation requirements. The embodiment of the present application does not limit the recommendation probability threshold and the position.

[0152] In this embodiment, the user product feature data set is predicted and processed through the product recommendation model to obtain target recommended products for target users, thereby realizing automatic determination of target recommended products, avoiding human participation, and improving the accuracy of target recommended products.

[0153] In an exemplary embodiment, Figure 7 As shown, the specific processing process in step 108 includes steps 702 to 704. Among them:

[0154] Step 702, obtaining the recommendation time of the target recommended product, and determining the target recommendation channel corresponding to the target recommended product in each channel.

[0155] In implementation, the computer device obtains a recommendation strategy template. The recommendation strategy includes various channels. The channel is a channel for sending the target recommendation strategy to the second business personnel. The computer device obtains the recommendation time of the target recommended product according to the recommendation strategy template, and determines the target recommendation channel corresponding to the target recommended product in each channel.

[0156] In an exemplary embodiment, the computer device obtains a recommendation strategy template, and updates the target recommendation product and the target recommendation words corresponding to the target recommendation product into the recommendation strategy template, and obtains the initial product recommendation strategy corresponding to the target recommendation product. The computer device displays the initial product recommendation strategy corresponding to the target recommendation product through the recommendation strategy webpage. The computer device obtains the recommendation time of the target recommendation product based on the initial product recommendation strategy. The initial product recommendation strategy includes various channels. The computer device determines the target recommendation channel corresponding to the target recommendation product in each recommendation channel.

[0157] In an optional embodiment, the target user inputs a recommendation time for the target recommended product based on the initial product recommendation strategy. The computer device obtains the recommendation time for the target recommended product input by the target user. The target user determines and triggers the target recommendation channel in each channel. The computer device obtains the target recommendation channel in response to the triggering operation of the target recommendation channel.

[0158] Step 704, construct a product recommendation strategy for the target recommended product of the target user based on the user feature data set of the target user, the target recommendation channel, the recommendation time, the second product attribute data subset corresponding to the target recommended product, and the target recommendation words.

[0159] During implementation, the computer device adds the target user's feature data set, target recommendation channel, recommendation time, and a second product attribute data subset corresponding to the target recommended product to the initial product recommendation strategy to obtain a product recommendation strategy for the target recommended product.

[0160] In an optional embodiment, the computer device adds the target user's feature data set, target recommendation channel, recommendation time, a second product attribute data subset corresponding to the target recommended product, and the target recommendation script to the recommendation strategy template to obtain a product recommendation strategy for the target recommended product.

[0161] In an exemplary embodiment, the computer device converts the product recommendation strategy into an information format acceptable to the second business personnel (such as personalized emails, SMS content) based on the target recommendation channel, and sends the converted product recommendation strategy to the second business personnel according to the recommendation time and the target recommendation channel, so that the second business personnel can receive the product recommendation strategy at the best time of the recommendation time and the most suitable channel of the target recommendation channel, and recommend the target recommended product to the target user based on the product recommendation strategy. It should be clear that the computer device achieves seamless integration with various channels through an API interface or a message queue (such as Kafka, an open source distributed stream processing platform). The computer device monitors the behavioral feedback of the target user through the terminal to monitor the customer's behavioral feedback in real time, such as click-through rate, purchase behavior, reply, etc. Then, the terminal returns these data to the computer device for analysis to obtain the analysis results. The implementation of the product recommendation strategy is further optimized through the analysis results, and the response rate and success rate of the product recommendation strategy are improved.

[0162] Optionally, the channel includes but is not limited to email, text message, telephone, social media, etc., and the embodiments of the present application do not limit the channel.

[0163] In this embodiment, a product recommendation strategy for the target recommended product of the target user is automatically generated through the target recommended product and the target recommended speech corresponding to the target recommended product, so as to realize the personalized generation of the product recommendation strategy, improve the acceptance of the target user and the conversion rate of the target recommended product. At the same time, by automatically generating the product recommendation strategy, the generation efficiency of the product recommendation strategy is improved.

[0164] In an exemplary embodiment, Figure 8 FIG. 1 is a schematic diagram of the architecture of a policy generation system in an embodiment. Figure 8 As shown, the strategy generation system includes a data collection and preprocessing module, a marketing strategy generation module, an automated marketing implementation module, an AI (Artificial Intelligence) model training and optimization module, and a feedback and strategy adjustment module. Among them:

[0165] The data collection and preprocessing module is used to obtain the user product feature data set, business personnel attribute data set, speech data set, product attribute data set and the user feature data set of the target user.

[0166] The AI ​​model training and optimization module is used to train the product recommendation model and language model. It is also used to determine the target recommended products corresponding to the target user based on the user product feature data set and the product recommendation model; to construct a product speech feature data set for each target recommended product based on the business personnel attribute data set, speech data set, product attribute data set and the user feature data set; and to determine the target recommended speech corresponding to the target recommended product based on the language model and the product speech feature data set.

[0167] The marketing strategy generation module is used to generate a product recommendation strategy for the target recommended product for the target user based on the target recommended product and the target recommendation speech corresponding to the target recommended product.

[0168] The automated marketing implementation module is used to send the product recommendation strategy of the target recommended product to the second business personnel through the target recommendation channel.

[0169] The main function of the feedback and strategy adjustment module is to adjust and optimize the product recommendation strategy in real time based on the feedback data of business personnel, and re-implement the adjusted product recommendation strategy. This module is also responsible for continuously optimizing the product recommendation model and language model to continuously evolve the strategy generation system.

[0170] Specifically, the feedback and strategy adjustment module collects user behavior data, product recommendation strategy response data and other feedback data in real time through message queues (such as RabbitMQ) or log analysis tools, and aggregates and analyzes the collected feedback data to extract effective information and obtain analysis results. Then, the feedback and strategy adjustment module adjusts the product recommendation strategy based on the analysis results of the feedback data, such as changing the recommended products, adjusting the marketing time, and modifying the communication content. The feedback and strategy adjustment module regenerates the optimized product recommendation strategy and automatically pushes it to business personnel. The feedback and strategy adjustment module inputs the feedback data into the product recommendation model and language model for retraining to improve the prediction accuracy of the product recommendation model and language model. The feedback and strategy adjustment module continuously monitors the performance of the product recommendation model and language model, and updates the model parameters or architecture as needed. The feedback and strategy adjustment module realizes the dynamic adjustment and continuous optimization of the product recommendation strategy, enabling the system to quickly adapt to user needs and maintain the efficiency and flexibility of the product recommendation strategy.

[0171] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0172] Based on the same inventive concept, the embodiment of the present application also provides a policy generation device for implementing the policy generation method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more policy generation device embodiments provided below can refer to the limitations on the policy generation method above, and will not be repeated here.

[0173] In an exemplary embodiment, Fig. 9 As shown, a strategy generation device 900 is provided, comprising: an acquisition module 901, a construction module 902, a determination module 903 and a generation module 904, wherein:

[0174] The acquisition module 901 is used to acquire the user product feature data set of the target user, and determine the target recommended products corresponding to the target user according to the user product feature data set and the product recommendation model; the user product feature data set includes the user feature data set of the target user.

[0175] The construction module 902 is used to construct a product speech feature data set for each target recommended product based on the business personnel attribute data set, the speech data set, the product attribute data set and the user feature data set.

[0176] The determination module 903 is used to determine the target recommendation speech corresponding to the target recommendation product based on the language model and the product speech feature data set.

[0177] The generating module 904 is used to generate a product recommendation strategy for the target recommended product for the target user according to the target recommended product and the target recommendation words corresponding to the target recommended product.

[0178] In an exemplary embodiment, the building block 902 includes:

[0179] The first determination submodule is used to determine, for each target recommended product, a second product attribute data subset of the target recommended product in the product attribute data set, and to determine each to-be-recommended speech corresponding to the target recommended product in the speech data set.

[0180] The second determination submodule is used to determine a second business personnel attribute data subset corresponding to the target user in the business personnel attribute data set.

[0181] The first construction submodule is used to construct a product speech feature data set of a target recommended product based on each speech to be recommended, the second product attribute data subset, the user feature data set and the second business personnel attribute data subset.

[0182] In an exemplary embodiment, the acquisition module 901 includes a first acquisition submodule and a third determination submodule. The first acquisition submodule includes:

[0183] The second acquisition submodule is used to obtain a user feature data set of the target user; the user feature data set includes a user attribute data set, a historical purchase record data set and an interest tag set of the target user; the historical purchase record data set includes a subset of historical purchase record data of each historically purchased product.

[0184] The fourth determination submodule is used to determine, for each historically purchased product, a first business person attribute data subset of the first business person in the historical purchase record data subset of the historically purchased product, in the business person attribute data set.

[0185] The fifth determination submodule is used to determine a first product attribute data subset of historically purchased products in the product attribute data set.

[0186] The second construction submodule is used to construct a user product feature data set for the target user based on the user feature data set, each first product attribute data subset and each first business personnel attribute data subset.

[0187] In an exemplary embodiment, the second acquisition submodule includes:

[0188] The third acquisition submodule is used to obtain the user attribute data set, the historical purchase record data set and the initial interest tag set of the target user.

[0189] The fourth acquisition submodule is used to acquire a social media tag data set, and update the initial interest tag set according to the social media tag data set to obtain an interest tag set.

[0190] The third construction submodule is used to construct a user feature dataset of the target user based on the interest tag set, the user attribute dataset and the historical purchase record dataset.

[0191] In an exemplary embodiment, the fourth acquisition submodule includes a fifth acquisition submodule and a first update submodule. The first update submodule includes:

[0192] The first judgment submodule is used to judge whether there is a target interest tag in the initial interest tag set that is the same as the target social media tag in the social media tag dataset.

[0193] The sixth determination submodule is used to determine, if there is a target interest tag that is the same as the target social media tag, an associated social media tag associated with the target social media tag in the social media tag dataset.

[0194] The adding submodule is used to add the associated social media tags to the initial interest tag set to obtain the interest tag set.

[0195] In an exemplary embodiment, the acquisition module 901 includes a first acquisition submodule and a third determination submodule. The third determination submodule includes:

[0196] The first processing submodule is used to perform prediction processing on the user product feature data set and the product list based on the product recommendation model to obtain the recommendation probability corresponding to each product in the product list.

[0197] The screening submodule is used to screen target recommended products from among the products based on the recommendation probability of each product and preset recommendation conditions.

[0198] In an exemplary embodiment, the generating module 904 includes:

[0199] The seventh determination submodule is used to obtain the recommendation time of the target recommended product and determine the target recommendation channel corresponding to the target recommended product in each channel.

[0200] The fourth construction submodule is used to construct a product recommendation strategy for the target recommended product of the target user based on the user feature data set of the target user, the target recommendation channel, the recommendation time, the second product attribute data subset corresponding to the target recommended product, and the target recommendation words.

[0201] Each module in the above-mentioned strategy generation device can be implemented in whole or in part by software, hardware and a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module above.

[0202] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig.10 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a strategy generation method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0203] Those skilled in the art will understand that Fig.10The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0204] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0205] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0206] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0207] 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 regulations.

[0208] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0209] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0210] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A strategy generation method, characterized in that: The method comprises: Acquire a user product feature data set of a target user, and determine each target recommended product corresponding to the target user according to the user product feature data set and a product recommendation model; the user product feature data set includes the user feature data set of the target user; Constructing a product speech feature data set for each target recommended product based on the business personnel attribute data set, the speech data set, the product attribute data set and the user feature data set; Determine the target recommendation speech corresponding to the target recommended product based on the language model and the product speech feature data set; A product recommendation strategy for the target recommended product for the target user is generated according to the target recommended product and the target recommendation speech corresponding to the target recommended product.

2. The method according to claim 1, characterized in that The step of constructing a product speech feature dataset for each target recommended product based on the business personnel attribute dataset, the speech dataset, the product attribute dataset and the user feature dataset includes: For each of the target recommended products, determine a second product attribute data subset of the target recommended product in the product attribute data set, and determine each to-be-recommended speech corresponding to the target recommended product in the speech data set; Determining a second business personnel attribute data subset corresponding to the target user in the business personnel attribute data set; A product speech feature data set of the target recommended product is constructed based on each of the to-be-recommended speech, the second product attribute data subset, the user feature data set and the second business personnel attribute data subset.

3. The method according to claim 1, characterized in that The step of obtaining a user product feature dataset of a target user includes: Acquire a user feature data set of a target user; the user feature data set includes a user attribute data set, a historical purchase record data set, and an interest tag set of the target user; the historical purchase record data set includes a historical purchase record data subset of each historically purchased product; For each of the historically purchased products, according to the first business person in the historical purchase record data subset of the historically purchased product, determining a first business person attribute data subset of the first business person in the business person attribute data set; Determining a first product attribute data subset of the historically purchased product in the product attribute data set; A user product feature data set of the target user is constructed based on the user feature data set, each of the first product attribute data subsets and each of the first business personnel attribute data subsets.

4. The method according to claim 3, characterized in that The step of obtaining a user feature data set of a target user includes: Obtain the user attribute dataset, historical purchase record dataset, and initial interest tag set of the target user; Acquire a social media tag data set, and update the initial interest tag set according to the social media tag data set to obtain an interest tag set; A user feature dataset of the target user is constructed according to the interest tag set, the user attribute dataset and the historical purchase record dataset.

5. The method according to claim 4, characterized in that The updating of the initial interest tag set according to the social media tag data set to obtain the interest tag set includes: Determine whether there is a target interest tag in the initial interest tag set that is the same as a target social media tag in the social media tag dataset; If the target interest tag is the same as the target social media tag, determining an associated social media tag associated with the target social media tag in the social media tag dataset; The associated social media tags are added to the initial interest tag set to obtain an interest tag set.

6. The method according to claim 1, characterized in that Determining each target recommended product corresponding to the target user according to the user product feature data set and the product recommendation model includes: Based on the product recommendation model, predictive processing is performed on the user product feature data set and the product list to obtain a recommendation probability corresponding to each product in the product list; Target recommended products are screened from among the products based on the recommendation probabilities of the products and preset recommendation conditions.

7. The method according to claim 1, characterized in that The generating a product recommendation strategy for the target recommended product for the target user according to the target recommended product and the target recommendation words corresponding to the target recommended product includes: Obtaining the recommendation time of the target recommended product, and determining the target recommendation channel corresponding to the target recommended product in each channel; A product recommendation strategy for the target recommended product of the target user is constructed based on the user feature data set of the target user, the target recommendation channel, the recommendation time, the second product attribute data subset corresponding to the target recommended product and the target recommendation words.

8. A strategy generation device, characterized in that: The device comprises: An acquisition module is used to acquire a user product feature data set of a target user, and determine each target recommended product corresponding to the target user according to the user product feature data set and a product recommendation model; the user product feature data set includes the user feature data set of the target user; A construction module, used to construct a product speech feature data set for each target recommended product based on the business personnel attribute data set, the speech data set, the product attribute data set and the user feature data set; A determination module, configured to determine a target recommendation speech corresponding to the target recommended product based on a language model and the product speech feature data set; A generation module is used to generate a product recommendation strategy for the target recommended product for the target user based on the target recommended product and the target recommendation speech corresponding to the target recommended product.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images