Product recommendation methods, product recommendation devices, electronic devices and storage media

By extracting features and analyzing intent from user data, a personalized product recommendation list is generated, which solves the problem of insufficient recommendation accuracy in existing technologies and achieves higher product recommendation accuracy.

CN116467517BActive Publication Date: 2026-11-17PING AN TECH (SHENZHEN) CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202310430243.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-04-12
Publication Date
2026-11-17
Estimated Expiration
2043-04-12

AI Technical Summary

Technical Problem

Existing product recommendation methods often fail to meet users' actual needs, resulting in insufficient recommendation accuracy.

Method used

By acquiring basic user data and user behavior data of target users, feature extraction is performed. A pre-set product recommendation model is used for product recall and intent data analysis. Based on the target intent features, the initial recommendation list is rearranged to generate a personalized target recommendation list.

Benefits of technology

This improves the accuracy of product recommendations and better meets users' personalized needs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116467517B_ABST
    Figure CN116467517B_ABST
Patent Text Reader

Abstract

This application provides a product recommendation method, a product recommendation device, an electronic device, and a storage medium, belonging to the field of artificial intelligence technology. The method includes: acquiring target user data, which includes the target user's basic user data and user behavior data; extracting features from the target user data to obtain target user features; recalling products based on a preset product recommendation model and the target user features to obtain an initial recommendation list; pushing the initial recommendation list to the target user and acquiring the intent data fed back by the target user based on the initial recommendation list; extracting entity features from the intent data to obtain target intent features; rearranging the products to be recommended in the initial recommendation list based on the target intent features to obtain a target recommendation list; and pushing the target recommendation list to the target user. This application embodiment can improve the accuracy of product recommendations.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a product recommendation method, a product recommendation device, an electronic device, and a storage medium. Background Technology

[0002] Current product recommendation methods often tend to recommend popular products to users. This approach often fails to meet users' actual needs and affects the accuracy of product recommendations. Therefore, improving the accuracy of product recommendations has become an urgent technical problem to be solved. Summary of the Invention

[0003] The main objective of this application is to provide a product recommendation method, a product recommendation device, an electronic device, and a storage medium, with the aim of improving the accuracy of product recommendations.

[0004] To achieve the above objectives, a first aspect of this application proposes a product recommendation method, the method comprising:

[0005] Acquire target user data, which includes the target user's basic user data and user behavior data;

[0006] Feature extraction is performed on the target user data to obtain target user features;

[0007] Based on the preset product recommendation model and the target user characteristics, product recall is performed to obtain an initial recommendation list;

[0008] The initial recommendation list is pushed to the target user, and the intent data fed back by the target user based on the initial recommendation list is obtained;

[0009] Entity features are extracted from the intent data to obtain the target intent features;

[0010] Based on the target intent features, the products to be recommended in the initial recommendation list are rearranged to obtain the target recommendation list;

[0011] The target recommendation list is pushed to the target user.

[0012] In some embodiments, the step of extracting features from the target user data to obtain target user features includes:

[0013] Feature extraction is performed on the user's basic data to obtain the user's basic features;

[0014] Feature extraction is performed on the user behavior data to obtain user behavior features;

[0015] The target user features are obtained by fusing the user behavior features and the user basic features.

[0016] In some embodiments, the step of retrieving products based on a preset product recommendation model and the target user characteristics to obtain an initial recommendation list includes:

[0017] The target user features are input into the product recommendation model, which includes an input layer, an encoding layer, and a prediction layer.

[0018] The target user features are embedded based on the input layer to obtain a user feature embedding vector.

[0019] The user feature embedding vector is encoded based on the encoding layer to obtain the user feature encoded vector.

[0020] Based on the prediction layer, the pre-acquired product features, and the user feature encoding vector, product recall is performed to obtain the initial recommendation list, where the product features are the basic features of the products to be recommended.

[0021] In some embodiments, the step of performing product recall based on the prediction layer, pre-acquired product features, and the user feature encoding vector to obtain the initial recommendation list includes:

[0022] Based on the prediction layer, the product recommendation score is calculated using the product features and the user feature encoding vector to obtain the recommendation score for each product to be recommended.

[0023] Based on the recommendation score and the preset recommendation threshold, the products to be recommended are screened to obtain candidate products;

[0024] The candidate products are sorted according to the recommendation scores to obtain the initial recommendation list.

[0025] In some embodiments, the step of extracting entity features from the intent data to obtain target intent features includes:

[0026] Traverse the preset intent database, calculate the correlation between the intent data and the reference intent features in the intent database, and obtain the intent relevance.

[0027] The target intent features are obtained by filtering the reference intent features based on the intent relevance.

[0028] In some embodiments, the step of reordering the products to be recommended in the initial recommendation list based on the target intent features to obtain the target recommendation list includes:

[0029] Based on the target intent features, calculate the predicted intent data for each product to be recommended in the initial recommendation list;

[0030] Based on the predicted intention data, the products to be recommended in the initial recommendation list are rearranged to obtain the target recommendation list.

[0031] In some embodiments, calculating the predicted intention data for each product to be recommended in the initial recommendation list based on the target intent features includes:

[0032] Based on the preset function and the target intent features, calculate the intent probability vector for each product to be recommended in the initial recommendation list;

[0033] The predicted intention data is obtained based on the intention probability vector.

[0034] To achieve the above objectives, a second aspect of this application provides a product recommendation device, the device comprising:

[0035] The data acquisition module is used to acquire target user data, which includes the target user's basic user data and user behavior data.

[0036] The feature extraction module is used to extract features from the target user data to obtain target user features;

[0037] The product recall module is used to recall products based on a preset product recommendation model and the target user characteristics to obtain an initial recommendation list.

[0038] The initial recommendation module is used to push the initial recommendation list to the target user and obtain the intent data fed back by the target user based on the initial recommendation list;

[0039] An entity feature extraction module is used to extract entity features from the intent data to obtain target intent features;

[0040] The rearrangement module is used to rearrange the products to be recommended in the initial recommendation list based on the target intent features to obtain the target recommendation list;

[0041] The target recommendation module is used to push the target recommendation list to the target user.

[0042] To achieve the above objectives, a third aspect of the present application provides an electronic device, the electronic device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method described in the first aspect.

[0043] To achieve the above objectives, a fourth aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect.

[0044] The product recommendation method, device, electronic equipment, and storage medium proposed in this application acquire target user data, including basic user data and user behavior data. Feature extraction is performed on the target user data to obtain target user features. This method conveniently obtains target user features containing important user information, enabling the use of these features in subsequent recommendation processes. Based on a pre-defined product recommendation model and target user features, product recall is performed to obtain an initial recommendation list. This allows for convenient filtering and sorting of products based on target user features, resulting in an initial recommendation list that matches the target user's user characteristics. The initial recommendation list is pushed to the target user, and the intent data provided by the target user based on the initial recommendation list is obtained. Entity feature extraction is performed on the intent data to obtain target intent features, which conveniently identifies the target user's current intent, enabling product recommendations based on the target user's current intent. Furthermore, based on the target intent features, the products in the initial recommendation list are rearranged to obtain a target recommendation list. This allows for convenient filtering and sorting of products based on the target user's target intent features, resulting in a target recommendation list that matches the target user's actual intent. Finally, the target recommendation list is pushed to the target user. This method can filter and sort the recommended products based on the user's characteristics and intent, and obtain a target recommendation list that matches the user's characteristics and current intent. This can better achieve personalized product recommendations and improve the accuracy of product recommendations. Attached Figure Description

[0045] Figure 1 This is a flowchart of the product recommendation method provided in the embodiments of this application;

[0046] Figure 2 yes Figure 1 The flowchart of step S102 in the document;

[0047] Figure 3 yes Figure 1 The flowchart of step S103 in the process;

[0048] Figure 4 yes Figure 3 The flowchart of step S304 in the process;

[0049] Figure 5 yes Figure 1 The flowchart of step S105 in the process;

[0050] Figure 6 yes Figure 1 The flowchart of step S106 in the process;

[0051] Figure 7 yes Figure 6 The flowchart of step S601 in the process;

[0052] Figure 8 This is a schematic diagram of the product recommendation device provided in the embodiments of this application;

[0053] Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0055] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0057] First, let's analyze some of the terms used in this application:

[0058] Artificial intelligence (AI) is a new branch of computer science that studies, develops, and applies theories, methods, technologies, and systems to simulate, extend, and expand human intelligence. It aims to understand the essence of intelligence and produce intelligent machines that can react in a way similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. AI can simulate the information processes of human consciousness and thought. Furthermore, AI utilizes digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceiving the environment, acquiring knowledge, and using that knowledge to achieve optimal results.

[0059] Natural Language Processing (NLP): NLP uses computers to process, understand, and utilize human language (such as Chinese and English). NLP is a branch of artificial intelligence and an interdisciplinary field of computer science and linguistics, often referred to as computational linguistics. NLP includes syntactic analysis, semantic analysis, and discourse understanding. It is commonly used in machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, intent recognition, information extraction and filtering, text classification and clustering, sentiment analysis, and opinion mining. It involves data mining, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research, and linguistic research related to language computation.

[0060] Information Extraction (NER) is a text processing technique that extracts factual information such as entities, relationships, and events from natural language text and outputs it as structured data. Information extraction is a technique for extracting specific information from text data. Text data is composed of specific units, such as sentences, paragraphs, and chapters. Text information is composed of smaller, specific units, such as characters, words, phrases, sentences, paragraphs, or combinations of these units. Extracting noun phrases, names of people, and place names from text data is an example of text information extraction. Of course, text information extraction techniques can extract information of various types.

[0061] Recall: Recall is the first stage of the recommendation system. It mainly uses some characteristics of users and products to quickly find a portion of items that users may be interested in from a massive item database, and then hands them over to the ranking stage.

[0062] An entity is a distinguishable and independently existing thing, such as a person, a city, a plant, or a commodity. Everything in the world is composed of concrete things, which are referred to here as entities. Entities are the most basic elements in a knowledge graph, and different entities have different relationships.

[0063] Embedding: Embedding is a vector representation that uses a low-dimensional vector to represent an object. This object could be a word, a product, a movie, etc. The property of this embedding vector is that vectors that are close in distance correspond to objects with similar meanings. Essentially, embedding is a mapping from semantic space to vector space, while preserving as much of the original sample's semantic relationship as possible in the vector space. For example, two semantically similar words are also likely to be located close in the vector space. Embedding can encode objects using low-dimensional vectors while retaining their meaning. It is commonly used in machine learning. In the process of building machine learning models, objects are encoded into low-dimensional dense vectors before being fed into a deep neural network (DNN) to improve efficiency.

[0064] Encoding: This involves transforming an input sequence into a vector of fixed length.

[0065] BERT (Bidirectional Encoder Representation from Transformers) model: The BERT model further enhances the generalization ability of word embedding models, fully describing character-level, word-level, sentence-level, and even inter-sentence relationship features, and is built based on Transformers. BERT has three types of embeddings: Token Embedding, Segment Embedding, and Position Embedding. Token Embeddings are word vectors, with the first word as a CLS marker, which can be used for subsequent classification tasks. Segment Embeddings are used to distinguish between two types of sentences because pre-training involves not only LM but also classification tasks with two sentences as input. Position Embeddings, where the positional word vectors are not trigonometric functions from Transformer, but rather learned by BERT during training. However, BERT directly trains a position embedding to preserve positional information. A vector is randomly initialized for each position and added to the model training, resulting in an embedding containing positional information. Finally, BERT chooses to directly concatenate this position embedding with the word embedding.

[0066] Current product recommendation methods often tend to recommend popular products to users. This approach often fails to meet users' actual needs and affects the accuracy of product recommendations. Therefore, improving the accuracy of product recommendations has become an urgent technical problem to be solved.

[0067] Based on this, embodiments of this application provide a product recommendation method, a product recommendation device, an electronic device, and a storage medium, aiming to improve the accuracy of product recommendations.

[0068] The product recommendation method, product recommendation device, electronic device, and storage medium provided in this application are specifically described through the following embodiments. First, the product recommendation method in this application embodiment is described.

[0069] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0070] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0071] The product recommendation method provided in this application relates to the field of artificial intelligence technology. The product recommendation method provided in this application can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the product recommendation method, but is not limited to the above forms.

[0072] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0073] Figure 1 This is an optional flowchart of the product recommendation method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S101 to S107.

[0074] Step S101: Obtain target user data, which includes the target user's basic user data and user behavior data;

[0075] Step S102: Extract features from the target user data to obtain target user features;

[0076] Step S103: Based on the preset product recommendation model and target user characteristics, product recall is performed to obtain an initial recommendation list;

[0077] Step S104: Push the initial recommendation list to the target user and obtain the intent data of the target user based on the initial recommendation list;

[0078] Step S105: Extract entity features from the intent data to obtain target intent features;

[0079] Step S106: Based on the target intent features, the products to be recommended in the initial recommendation list are rearranged to obtain the target recommendation list;

[0080] Step S107: Push the target recommendation list to the target user.

[0081] Steps S101 to S107 of this embodiment involve acquiring target user data, including basic user data and user behavior data; extracting features from the target user data to obtain target user features, which conveniently yields target user features containing important user information, enabling their use in subsequent recommendation processes. Based on a preset product recommendation model and target user features, product recall is performed to obtain an initial recommendation list, which facilitates the filtering and sorting of recommended products based on target user features, resulting in an initial recommendation list that matches the target user's user characteristics. The initial recommendation list is pushed to the target user, and the intent data fed back by the target user based on the initial recommendation list is obtained; entity feature extraction is performed on the intent data to obtain target intent features, which conveniently identifies the target user's current intent, enabling product recommendations based on the target user's current intent. Furthermore, based on the target intent features, the products to be recommended in the initial recommendation list are rearranged to obtain a target recommendation list, which facilitates the filtering and sorting of recommended products based on the target user's target intent features, resulting in a target recommendation list that matches the target user's actual intent. Finally, the target recommendation list is pushed to the target user. This method can filter and sort the recommended products based on the user's characteristics and intent, and obtain a target recommendation list that matches the user's characteristics and current intent. This can better achieve personalized product recommendations and improve the accuracy of product recommendations.

[0082] In step S101 of some embodiments, target user data can be obtained by web crawling. For example, the target user data can be obtained by crawling a preset data source. The preset data source may include a user database or a network platform commonly used by the user. The target user data includes the target user's basic user data and user behavior data. The basic user data includes the user's age, gender, education level, etc., and the user behavior data includes the target user's activity level, potential tags, click records, access records, purchase records, etc.

[0083] Please see Figure 2 In some embodiments, step S102 may include, but is not limited to, steps S201 to S203:

[0084] Step S201: Extract features from user basic data to obtain user basic features;

[0085] Step S202: Extract features from user behavior data to obtain user behavior features;

[0086] Step S203: Perform feature fusion on user behavior features and user basic features to obtain target user features.

[0087] In step S201 of some embodiments, user basic data can be feature extracted using a preset BERT model or a named entity extraction algorithm to obtain user basic features. Taking the BERT model as an example, user basic data is input into a pre-trained BERT model. The BERT model performs sentence segmentation on the user basic data to obtain multiple user basic word segments. Then, based on a preset dictionary, the user basic word segments are transformed. According to the mapping relationship between numbers and words in the preset dictionary, the user basic word segments are converted into corresponding numbers. Then, the encoder in the BERT model encodes the numbers corresponding to the user basic word segments to obtain user basic features. These user basic features include feature information such as the target user's gender and age.

[0088] In step S202 of some embodiments, user behavior data can be feature extracted using a preset BERT model or a named entity extraction algorithm to obtain user behavior features. Taking the BERT model as an example, user behavior data is input into a pre-trained BERT model. The BERT model performs sentence segmentation on the user behavior data to obtain multiple user behavior segments. Then, based on a preset dictionary, the user behavior segments are transformed into corresponding numbers according to the mapping relationship between numbers and words in the preset dictionary. Finally, the encoder in the BERT model encodes the numbers corresponding to the user behavior segments to obtain user behavior features. These user behavior features include feature information such as the target user's click records, purchase records, and browsing records.

[0089] In step S203 of some embodiments, user behavior features and user basic features can be converted into vector form, and then the vector form of user behavior features and user basic features can be concatenated to obtain target user features, wherein the target user features include user basic feature information and user behavior feature information.

[0090] Steps S201 to S203 above extract features from user basic data and user behavior data, which can conveniently obtain user basic features and user behavior features that characterize the target user. The user basic features and user behavior features are then fused to obtain target user features containing important user information, which can be used in the subsequent recommendation process to improve recommendation accuracy.

[0091] Please see Figure 3 In some embodiments, step S103 may include, but is not limited to, steps S301 to S304:

[0092] Step S301: Input the target user features into the product recommendation model. The product recommendation model includes an input layer, an encoding layer, and a prediction layer.

[0093] Step S302: Based on the input layer, the target user features are embedded to obtain the user feature embedding vector;

[0094] Step S303: Encode the user feature embedding vector based on the coding layer to obtain the user feature coding vector;

[0095] Step S304: Based on the prediction layer, pre-acquired product features, and user feature encoding vectors, product recall is performed to obtain an initial recommendation list, where the product features are the basic features of the products to be recommended.

[0096] In step S301 of some embodiments, the preset product recommendation model can be constructed based on deep learning neural networks or random forest algorithms, etc. The product recommendation model includes an input layer, an encoding layer, and a prediction layer. The input layer is used to embed the input features and map the input features to a fixed-dimensional vector space. The encoding layer is used to encode the input features in the fixed vector space to obtain important feature information and obtain feature encoding vectors. The prediction layer is used to make recommendation predictions based on the feature encoding vectors to obtain a recommendation score for each candidate product and generate a recommendation list based on the recommendation scores. Furthermore, this product recommendation model can be aggregated from recommendation models of various product types. For example, specific recommendation models can be constructed for various products in the financial field, such as deposits, fixed deposits, current accounts, public funds, and private funds. Then, a general recommendation model can be constructed based on the common features of these products. Multiple specific recommendation models and the general recommendation model are aggregated according to preset model weights to obtain the aforementioned product recommendation model. The model structures of the specific and general recommendation models can be consistent with the product recommendation model described above; that is, the specific and general recommendation models can also be constructed based on deep learning neural networks or random forest algorithms. Both the specific and general recommendation models include an input layer, an encoding layer, and a prediction layer. During model training, the specific recommendation models for different products are trained based on the unique product features of that product. For example, the products to be recommended can be divided into wealth management products, fund products, etc. The unique product features of wealth management products include the minimum investment amount, the proportion of equity assets, and the proportion of debt assets; the unique product features of fund products include the drawdown rate, the fund company, and the fund manager, etc. Similarly, a general recommendation model can be trained based on common product features across different products. These common features include, but are not limited to, product type, product yield, product duration, product exposure, and product browsing activity. Furthermore, the commonly used cross-entropy function can be used as the loss function during model training.

[0097] It should be noted that when inputting target user characteristics into the product recommendation model, the target user characteristics can be input into the product recommendation model through preset code programs or script programs.

[0098] In step S302 of some embodiments, the target user features are embedded through the input layer, and the target user features are mapped to a vector space of fixed dimensions to obtain user feature embedding vectors. The dimension of the vector space can be set according to actual business needs and is not limited.

[0099] In step S303 of some embodiments, the user feature embedding vector is encoded by an encoding layer to obtain more important feature information in the user feature embedding vector, thereby obtaining the user feature encoding vector.

[0100] In step S304 of some embodiments, product recommendation scores are calculated based on the product feature and user feature encoding vectors of the prediction layer to obtain the recommendation score of each product to be recommended. Then, the recommendation score is compared with a preset recommendation threshold to filter out products to be recommended with recommendation scores higher than the recommendation threshold, thus obtaining candidate products. Then, according to the recommendation scores of the candidate products, a series of candidate products are sorted in descending order to obtain an initial recommendation list.

[0101] Through the above steps S301 to S304, it is relatively convenient to filter and sort the recommended products based on the characteristics of the target users, and obtain an initial recommendation list that matches the user characteristics of the target users. Different initial recommendation lists can be constructed for different target users, which improves the accuracy of recommendations and enables personalized product recommendations.

[0102] Please see Figure 4 In some embodiments, step S304 may include, but is not limited to, steps S401 to S403:

[0103] Step S401: Calculate the product recommendation score based on the product feature and user feature encoding vectors of the prediction layer to obtain the recommendation score for each product to be recommended.

[0104] Step S402: Based on the recommendation score and the preset recommendation threshold, the products to be recommended are screened to obtain candidate products;

[0105] Step S403: Sort the candidate products according to the recommendation scores to obtain an initial recommendation list.

[0106] In step S401 of some embodiments, the product features are the basic features of the products to be recommended, including the product type, product name, product term, and basic product parameters, etc. The products to be recommended include wealth management products, fund products, etc. The product features are converted into vector form to obtain the product feature vector for each product to be recommended. The similarity between the product feature vector and the user feature encoding vector is calculated using collaborative filtering algorithms, alternating least squares, or dual-tower algorithms, etc., to obtain a vector similarity value. This vector similarity value is used as the recommendation score for each product to be recommended.

[0107] In step S402 of some embodiments, the preset recommendation threshold can be set according to actual needs and is not limited. The recommendation score is compared with the preset recommendation threshold. The higher the recommendation score, the more the recommended product meets the target user's current needs. Therefore, based on the relationship between the recommendation score and the recommendation threshold, several recommended products with recommendation scores higher than the recommendation threshold are selected from the list of recommended products. These selected recommended products are then used as candidate products. This series of candidate products represents the products that meet the target user's current needs.

[0108] In step S403 of some embodiments, the candidate products are arranged in descending order of recommendation score according to the recommendation score of each candidate product to obtain an initial recommendation list.

[0109] Through the above steps S401 to S403, it is relatively convenient to filter and sort the products to be recommended based on product features and target user features, obtain an initial recommendation list that matches the user features of the target user, and recommend products to the target user based on the initial recommendation list, so that the candidate products in the initial recommendation list can meet the current needs of the target user, thereby improving the accuracy of the recommendation.

[0110] In step S104 of some embodiments, an initial recommendation list is pushed to the target user via wireless or wired communication, enabling the target user to browse, click, or purchase candidate products in the initial recommendation list, generating user operation information. This user operation information can characterize the target user's interest and preferences for candidate products. For example, the target user can favorite candidate products they are interested in and mark candidate products they are not interested in. The target user can also input filtering conditions according to actual needs, such as "yield higher than a certain value". Therefore, user operation information is generated in response to the target user's operation behavior, and this user operation information includes the target user's intent data. Further, the intent data fed back by the target user based on the initial recommendation list is obtained via wireless or wired communication. This intent data can characterize the target user's intention for candidate products in the initial recommendation list.

[0111] Please see Figure 5 In some embodiments, step S105 may include, but is not limited to, steps S501 to S502:

[0112] Step S501: Traverse the preset intent database, calculate the correlation between the intent data and the reference intent features in the intent database, and obtain the intent relevance.

[0113] Step S502: Filter the reference intent features based on intent relevance to obtain the target intent features.

[0114] In step S501 of some embodiments, the preset intent database contains various reference intent features. For example, reference intent features include, but are not limited to, product type, investment direction, investment period, and investment rate of return. This intent database can be dynamically adjusted according to business needs. By traversing the preset intent database, the correlation between the intent data and the reference intent features in the intent database is calculated using collaborative filtering algorithms, etc., to obtain the intent relevance between the intent data and each reference intent feature. The intent relevance characterizes the degree of similarity between the intent data and the reference intent features.

[0115] In step S502 of some embodiments, since intent relevance can characterize the degree of similarity between intent data and reference intent features, if the intent relevance between intent data and a certain reference intent feature is higher, it indicates that the user intent in the intent data is closer to the reference intent feature. Therefore, the intent relevance of each reference intent feature is compared, and the reference intent feature with the highest intent relevance is selected as the target intent feature.

[0116] For example, the reference intent features include an investment period of 6 months, an investment period of 1 year, and an investment period of 3 years. The intent correlation between the investment period of 6 months and the intent data is 0.3, the intent correlation between the investment period of 1 year and the intent data is 0.6, and the intent correlation between the investment period of 3 years and the intent data is 0.05. This indicates that the target user tends to choose the product with an investment period of 1 year, and the reference intent feature of an investment period of 1 year is used as the target intent feature.

[0117] Through the above steps S501 to S502, reference intent features with high relevance to the intent data of the target user can be easily filtered from the preset intent database to obtain the target intent features. This makes it easier to identify the current intent of the target user, enabling product recommendations based on the current intent of the target user and effectively improving the accuracy of product recommendations.

[0118] Please see Figure 6 In some embodiments, step S106 includes, but is not limited to, steps S601 to S602:

[0119] Step S601: Calculate the predicted intention data for each product to be recommended in the initial recommendation list based on the target intent features;

[0120] Step S602: Based on the predicted intention data, the products to be recommended in the initial recommendation list are rearranged to obtain the target recommendation list.

[0121] In step S601 of some embodiments, the similarity between the target intent features and the product features of each product to be recommended in the initial recommendation list is calculated using collaborative filtering algorithm, alternating least squares method or dual-tower algorithm, etc., to obtain feature similarity value, and the feature similarity value is used as the predicted intent data of each product to be recommended.

[0122] In step S602 of some embodiments, the products to be recommended in the initial recommendation list are arranged in descending order of predicted intention data according to the predicted intention data of each product to be recommended in the initial recommendation list, so as to obtain the target recommendation list.

[0123] Through the above steps S601 to S602, it is relatively convenient to filter and sort the recommended products based on the target user's target intent characteristics, obtain a target recommendation list that matches the target user's actual intent, and recommend products to the target user according to the target recommendation list, so that the candidate products in the target recommendation list can meet the target user's current needs, thereby improving the accuracy of the recommendation.

[0124] Please see Figure 7 In some embodiments, step S601 may include, but is not limited to, steps S701 to S702:

[0125] Step S701: Based on the preset function and target intent features, calculate the intent probability vector of each product to be recommended in the initial recommendation list;

[0126] Step S702: Obtain the predicted intention data based on the intention probability vector.

[0127] In step S701 of some embodiments, the preset function can be a probability function such as the softmax function. Taking the softmax function as an example, the softmax function is used to create the probability distribution of the target intent feature on each product to be recommended, and the intent probability vector of each product to be recommended in the initial recommendation list is obtained. The intent probability vector is used to reflect the target user's intent tendency for each product to be recommended. That is, the larger the intent probability vector of a certain product to be recommended, the greater the target user's intent tendency for this product to be recommended.

[0128] In step S702 of some embodiments, the intent probability vector of each product to be recommended in the initial recommendation list is used as the predicted intent data of each product to be recommended.

[0129] Through the above steps S701 to S702, the intent probability vector of each product to be recommended in the initial recommendation list can be calculated relatively easily. Based on the intent probability vector, the target user's intention for each product to be recommended can be determined, so that products that match the target user's current intention can be selected for recommendation based on the target user's intention, thereby improving the accuracy of product recommendation.

[0130] In step S107 of some embodiments, the target recommendation list is directly pushed to the target user, or products that are at the top of the target recommendation list are selected and pushed to the target user, thereby reducing communication costs while achieving personalized recommendations.

[0131] The product recommendation method of this application embodiment obtains target user data, including the target user's basic user data and user behavior data; it extracts features from the target user data to obtain target user features, which can conveniently obtain target user features containing important user information, enabling the target user features to be used in subsequent recommendation processes. Based on a preset product recommendation model and target user features, product recall is performed to obtain an initial recommendation list, which can conveniently filter and sort products to be recommended based on target user features, resulting in an initial recommendation list that matches the target user's user characteristics. The initial recommendation list is pushed to the target user, and the intent data fed back by the target user based on the initial recommendation list is obtained; entity feature extraction is performed on the intent data to obtain target intent features, which can conveniently identify the target user's current intent, enabling product recommendations based on the target user's current intent. Furthermore, based on the target intent features, the products to be recommended in the initial recommendation list are rearranged to obtain a target recommendation list, which can conveniently filter and sort products to be recommended based on the target user's target intent features, resulting in a target recommendation list that matches the target user's actual intent. Finally, the target recommendation list is pushed to the target user. This method can filter and sort the recommended products based on the user's characteristics and intent, and obtain a target recommendation list that matches the user's characteristics and current intent. This can better achieve personalized product recommendations and improve the accuracy of product recommendations.

[0132] Please see Figure 8 This application also provides a product recommendation device that can implement the above-described product recommendation method. The device includes:

[0133] The data acquisition module 801 is used to acquire target user data, which includes the target user's basic user data and user behavior data.

[0134] The feature extraction module 802 is used to extract features from the target user data to obtain the target user features;

[0135] Product recall module 803 is used to recall products based on a preset product recommendation model and target user characteristics to obtain an initial recommendation list;

[0136] The initial recommendation module 804 is used to push the initial recommendation list to the target user and obtain the intent data of the target user based on the initial recommendation list;

[0137] The entity feature extraction module 805 is used to extract entity features from the intent data to obtain the target intent features;

[0138] The rearrangement module 806 is used to rearrange the products to be recommended in the initial recommendation list based on the target intent features to obtain the target recommendation list;

[0139] The target recommendation module 807 is used to push the target recommendation list to the target user.

[0140] In some embodiments, the feature extraction module 802 includes:

[0141] The first feature extraction unit is used to extract features from the user's basic data to obtain the user's basic features;

[0142] The second feature extraction unit is used to extract features from user behavior data to obtain user behavior features;

[0143] The feature fusion unit is used to fuse user behavior features and basic user features to obtain target user features.

[0144] In some embodiments, the product recall module 803 includes:

[0145] The input unit is used to input the target user features into the product recommendation model. The product recommendation model includes an input layer, an encoding layer, and a prediction layer.

[0146] The embedding unit is used to embed the target user features based on the input layer to obtain the user feature embedding vector;

[0147] The encoding unit is used to encode the user feature embedding vector based on the encoding layer to obtain the user feature encoded vector;

[0148] The recall unit is used to recall products based on the prediction layer, pre-acquired product features, and user feature encoding vectors to obtain an initial recommendation list. The product features are the basic features of the products to be recommended.

[0149] In some embodiments, the recall unit includes:

[0150] The score calculation subunit is used to calculate the product recommendation score based on the product feature and user feature encoding vector of the prediction layer, and to obtain the recommendation score of each product to be recommended.

[0151] The filtering subunit is used to filter the products to be recommended based on the recommendation score and the preset recommendation threshold to obtain candidate products;

[0152] The sorting subunit is used to sort candidate products according to their recommendation scores to obtain an initial recommendation list.

[0153] In some embodiments, the entity feature extraction module 805 includes:

[0154] The relevance calculation unit is used to traverse the preset intent database, perform relevance calculation on the intent data and the reference intent features in the intent database, and obtain the intent relevance.

[0155] The feature filtering unit is used to filter reference intent features based on intent relevance to obtain target intent features.

[0156] In some embodiments, the rearrangement module 806 includes:

[0157] The intent calculation unit is used to calculate the predicted intent data for each product to be recommended in the initial recommendation list based on the target intent characteristics.

[0158] The rearrangement unit is used to rearrange the products to be recommended in the initial recommendation list based on the predicted intention data, so as to obtain the target recommendation list.

[0159] In some embodiments, the intent calculation unit includes:

[0160] The probability calculation subunit is used to calculate the intent probability vector of each product to be recommended in the initial recommendation list based on a preset function and target intent features;

[0161] The data generation subunit is used to obtain predicted intention data based on the intention probability vector.

[0162] The specific implementation of the product recommendation device is basically the same as the specific implementation of the product recommendation method described above, and will not be repeated here.

[0163] This application also provides an electronic device, which includes: a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for communication between the processor and the memory. When the program is executed by the processor, it implements the product recommendation method described above. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0164] Please see Figure 9 , Figure 9 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:

[0165] The processor 901 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0166] The memory 902 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called and executed by the processor 901 using the product recommendation method of the embodiments of this application.

[0167] The 903 input / output interface is used to implement information input and output.

[0168] The communication interface 904 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0169] Bus 905 transmits information between various components of the device (e.g., processor 901, memory 902, input / output interface 903, and communication interface 904);

[0170] The processor 901, memory 902, input / output interface 903, and communication interface 904 are connected to each other within the device via bus 905.

[0171] This application also provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the product recommendation method described above.

[0172] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0173] The product recommendation method, device, electronic device, and computer-readable storage medium provided in this application acquire target user data, including basic user data and user behavior data. Feature extraction is performed on the target user data to obtain target user features. This method can conveniently obtain target user features containing important user information, enabling the use of these features in subsequent recommendation processes. Based on a preset product recommendation model and target user features, product recall is performed to obtain an initial recommendation list. This allows for convenient filtering and sorting of products to be recommended based on target user features, resulting in an initial recommendation list that matches the target user's user characteristics. The initial recommendation list is pushed to the target user, and the intent data fed back by the target user based on the initial recommendation list is obtained. Entity feature extraction is performed on the intent data to obtain target intent features, which can conveniently identify the target user's current intent, enabling product recommendations based on the target user's current intent. Furthermore, based on the target intent features, the products to be recommended in the initial recommendation list are rearranged to obtain a target recommendation list. This allows for convenient filtering and sorting of products to be recommended based on the target user's target intent features, resulting in a target recommendation list that matches the target user's actual intent. Finally, the target recommendation list is pushed to the target user. This method can filter and sort the recommended products based on the user's characteristics and intent, and obtain a target recommendation list that matches the user's characteristics and current intent. This can better achieve personalized product recommendations and improve the accuracy of product recommendations.

[0174] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0175] It will be understood by those skilled in the art that Figure 1-7 The technical solutions shown do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0176] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0177] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0178] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0179] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0180] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0181] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0182] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0183] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0184] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A product recommendation method, characterized in that, The method includes: Acquire target user data, which includes the target user's basic user data and user behavior data; Feature extraction is performed on the target user data to obtain target user features; Based on the preset product recommendation model and the target user characteristics, product recall is performed to obtain an initial recommendation list; The initial recommendation list is pushed to the target user, and the intent data fed back by the target user based on the initial recommendation list is obtained; Entity features are extracted from the intent data to obtain the target intent features; Based on the target intent features, the products to be recommended in the initial recommendation list are rearranged to obtain the target recommendation list; Push the target recommendation list to the target user; The step of extracting entity features from the intent data to obtain target intent features includes: Traverse the preset intent database, perform correlation calculation on the intent data and the reference intent features in the intent database to obtain the intent relevance, and select the reference intent feature with the highest intent relevance as the target intent feature. The step of extracting features from the target user data to obtain target user features includes: Feature extraction is performed on the user's basic data to obtain user basic features; feature extraction is performed on the user's behavior data to obtain user behavior features; feature fusion is performed on the user behavior features and the user's basic features to obtain the target user features. The user's basic data includes the target user's age, gender, and education level; the user behavior data includes the target user's activity level, potential tags, click records, access records, and purchase records.

2. The product recommendation method according to claim 1, characterized in that, The product recall based on the preset product recommendation model and the target user characteristics, to obtain an initial recommendation list, includes: The target user features are input into the product recommendation model, which includes an input layer, an encoding layer, and a prediction layer. The target user features are embedded based on the input layer to obtain a user feature embedding vector. The user feature embedding vector is encoded based on the encoding layer to obtain the user feature encoded vector. Based on the prediction layer, the pre-acquired product features, and the user feature encoding vector, product recall is performed to obtain the initial recommendation list, where the product features are the basic features of the products to be recommended.

3. The product recommendation method according to claim 2, characterized in that, The product recall based on the prediction layer, pre-acquired product features, and the user feature encoding vector, to obtain the initial recommendation list, includes: Based on the prediction layer, the product recommendation score is calculated using the product features and the user feature encoding vector to obtain the recommendation score for each product to be recommended. Based on the recommendation score and the preset recommendation threshold, the products to be recommended are screened to obtain candidate products; The candidate products are sorted according to the recommendation scores to obtain the initial recommendation list.

4. The product recommendation method according to claim 1, characterized in that, The step of reordering the products to be recommended in the initial recommendation list based on the target intent features to obtain the target recommendation list includes: Based on the target intent features, calculate the predicted intent data for each product to be recommended in the initial recommendation list; Based on the predicted intention data, the products to be recommended in the initial recommendation list are rearranged to obtain the target recommendation list.

5. The product recommendation method according to claim 4, characterized in that, The step of calculating the predicted intention data for each product to be recommended in the initial recommendation list based on the target intention features includes: Based on the preset function and the target intent features, calculate the intent probability vector for each product to be recommended in the initial recommendation list; The predicted intention data is obtained based on the intention probability vector.

6. A product recommendation device, characterized in that, The device includes: The data acquisition module is used to acquire target user data, which includes the target user's basic user data and user behavior data. The feature extraction module is used to extract features from the target user data to obtain target user features; The product recall module is used to recall products based on a preset product recommendation model and the target user characteristics to obtain an initial recommendation list. The initial recommendation module is used to push the initial recommendation list to the target user and obtain the intent data fed back by the target user based on the initial recommendation list; An entity feature extraction module is used to extract entity features from the intent data to obtain target intent features; The rearrangement module is used to rearrange the products to be recommended in the initial recommendation list based on the target intent features to obtain the target recommendation list; The target recommendation module is used to push the target recommendation list to the target user; The step of extracting entity features from the intent data to obtain target intent features includes: Traverse the preset intent database, perform correlation calculation on the intent data and the reference intent features in the intent database to obtain the intent relevance, and select the reference intent feature with the highest intent relevance as the target intent feature. The step of extracting features from the target user data to obtain target user features includes: Feature extraction is performed on the user's basic data to obtain user basic features; feature extraction is performed on the user's behavior data to obtain user behavior features; feature fusion is performed on the user behavior features and the user's basic features to obtain the target user features. The user's basic data includes the target user's age, gender, and education level; the user behavior data includes the target user's activity level, potential tags, click records, access records, and purchase records.

7. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the product recommendation method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the product recommendation method according to any one of claims 1 to 5.

Citation Information

Patent Citations

  • Product recommendation method, device and storage medium

    CN112785397A

  • User intention recognition method and device, electronic equipment and storage medium

    CN115270900A