Product recommendation method and device, equipment, storage medium and program product

Abstract

The invention relates to the field of precision marketing, and provides a product recommendation method, device and equipment, a storage medium and a program product.The method comprises the steps that a standardized demand corresponding to the purchase intention of a target user and historical sales data under the standardized demand are obtained; obtaining each to-be-recommended product under the standardized demand from a standardized product database to obtain a product recommendation pool; and according to the historical sales data, constructing a user portrait under a standardized demand, and based on the user portrait, matching a target product from a product recommendation pool and recommending the target product to a target user. According to the method, standardized precipitation is carried out on business scenes and demands, user portraits under standardized demands are matched with products based on association relationships among the scenes, the demands and the products, different product feature intentions corresponding to differentiated demands under different scenes are considered, the matching degree of the products and the demands is improved, and the user experience is improved. Therefore, the accuracy of product recommendation is improved, and meanwhile, business personnel can focus on user requirements to improve the business expansion efficiency.

Description

Technical Field

[0001] The present application relates to the field of precision marketing, and in particular to a product recommendation method, device, equipment, storage medium and program product. Background Art

[0002] In the process of business operation and development, with the continuous enrichment of business scenarios and products, business expansion often faces complex needs of users. The traditional business expansion model of "one-on-one discussion" and "single product sales" is inefficient and can no longer adapt to the high-quality development needs of enterprises. In order to meet user needs, improve sales efficiency and maintain competitive advantages, it is necessary to formulate and optimize product expansion strategies in a timely manner to achieve precision marketing.

[0003] At present, most product recommendation methods for complex needs are to build user portraits through user historical data, and use big data and related algorithms to recommend products to achieve precision marketing. The user portraits built by this method are relatively macro and general, and do not take into account the differentiated needs of users in different scenarios. For different needs, the product features required by users will also be different, which will lead to inaccurate matching between recommended products and user needs, affecting the accuracy of product recommendations and further affecting the order conversion rate. Summary of the invention

[0004] The present application provides a product recommendation method, device, equipment, storage medium and program product to solve the defects of the existing product recommendation method, such as low business expansion efficiency, inaccurate matching of recommended products with user needs, and affecting the accuracy of product recommendations.

[0005] The present application provides a product recommendation method, comprising: Obtaining standardized demands corresponding to the target user's purchase intentions and historical sales data under the standardized demands; Acquire each product to be recommended under the standardized demand from a standardized product database to obtain a product recommendation pool; the standardized product database is obtained by associating standardized demands and products under various standardized scenarios, and the standardized demands are obtained by performing natural language processing on demand documents under various standardized scenarios; A user profile under the standardized requirements is constructed according to the historical sales data, and a target product is matched from the product recommendation pool based on the user profile and recommended to the target user.

[0006] In one embodiment, constructing the user profile under the standardized requirement according to the historical sales data, and matching the target product from the product recommendation pool based on the user profile and recommending it to the target user includes: Determine the sold products under the standardized requirements according to the historical sales data; Based on preset influencing characteristic factors, extract the first product characteristic vector of the sold product, and extract the second product characteristic vector of each product to be recommended in the product recommendation pool; the influencing characteristic factors are the product characteristics that have the greatest impact on the user's order conversion rate, and the influencing characteristic factors include product price, product delivery time, and after-sales service response time; Calculating a user preference vector based on the first product feature vector to construct a user portrait; Calculating the similarity between the user preference vector corresponding to the user portrait and the second product feature vector of each of the products to be recommended; The products to be recommended are sorted in descending order according to the similarity, and a first preset number of products to be recommended are selected in the sorting order as target products to be recommended to the target user.

[0007] In one embodiment, before obtaining the standardized requirements corresponding to the target user's purchase intention, the method further includes: Obtaining full product data and demand documents for each preset standardized scenario; the full product data includes product description information of each product; Reading the demand text in the demand document, and performing word segmentation and stop word removal on the demand text and the product description information, respectively, to obtain a first word set corresponding to the demand document and a second word set corresponding to the product description information; Calculate a first word frequency feature value of each word in the demand text based on the first word set, and calculate a second word frequency feature value of each word in the product description information based on the second word set; Selecting a plurality of word combinations with the largest first word frequency characteristic value from the demand text as process keyword groups, and selecting a second preset number of words from the product description information as description keywords; Clustering and grouping the process keyword phrases, and selecting the words with the highest word frequency in each group as the standardized requirements, constructing a corresponding relationship between the standardized requirements and the standardized scenarios, and generating a standardized scenario requirements database; Performing text similarity analysis on the standardized requirements and the description keywords to determine the text similarity between the standardized requirements and each of the products; On the basis of the standardized scenario requirement database, the products whose text similarity is greater than a preset reference value are associated with the standardized requirements to obtain a standardized product database.

[0008] In one embodiment, the step of calculating a first word frequency feature value of each word in the demand text based on the first word set, and calculating a second word frequency feature value of each word in the product description information based on the second word set includes: Counting the first word frequency of a first target word in the first word set in a target document, and the first document quantity of a demand document containing the first target word; the first target word is any word in the first word set, and the target document is the demand document to which the first target word belongs; Calculating a first inverse document frequency of the first target term in the demand document according to the total number of documents in the demand document and the first number of documents; Calculate the product of the first word frequency and the first inverse document frequency to obtain a first word frequency feature value of the first target word; Counting a second word frequency of a second target word in the second word set in the target description information, and a first product quantity of the product containing the second target word in the product description information; the second target word is any word in the second word set, and the target description information is product description information of the product to which the second target word belongs; Calculating a second inverse document frequency of the second target word in the product description information according to the total product quantity of the product and the first product quantity; The product of the second word frequency and the second inverse document frequency is calculated to obtain a second word frequency feature value of the second target word.

[0009] In one embodiment, clustering and grouping the process keyword groups includes: Extracting a vector representation of each process keyword in the process keyword group; Calculate the cosine similarity between each of the process keywords based on the vector representation to obtain a similarity matrix corresponding to the process keyword group; Based on the similarity matrix and a preset similarity threshold, an agglomerative hierarchical clustering algorithm is used to cluster and group the process keyword phrases.

[0010] In one embodiment, after matching the target product from the product recommendation pool and recommending it to the target user based on the user portrait, the method further includes: Obtaining a selection operation for the target product; A product solution configuration sheet and a price list are generated according to the target product selected by the selection operation.

[0011] The present application also provides a product recommendation device, comprising the following modules: A demand acquisition module is used to acquire the standardized demand corresponding to the target user's purchase intention and the historical sales data under the standardized demand; A product acquisition module is used to obtain each product to be recommended under the standardized demand from a standardized product database to obtain a product recommendation pool; the standardized product database is obtained by associating standardized demands and products under various standardized scenarios, and the standardized demands are obtained by performing natural language processing on demand documents under various standardized scenarios; The product recommendation module is used to construct a user portrait under the standardized requirements according to the historical sales data, and match the target product from the product recommendation pool based on the user portrait and recommend it to the target user.

[0012] The present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the above-described product recommendation methods when executing the computer program.

[0013] The present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the product recommendation methods described above.

[0014] The present application also provides a computer program product, including a computer program, which implements the steps of any of the above-mentioned product recommendation methods when executed by a processor.

[0015] The product recommendation method, device, equipment, storage medium and program product provided by the present application are based on a standardized product database, and the standardized needs of users are determined by performing demand decomposition on the user's purchase intention, and the products to be recommended under the standardized needs of users are obtained from the standardized product database to form a product recommendation pool, and then a user portrait is constructed according to the historical sales data under the standardized needs, and the target products are matched from the product recommendation pool and recommended to the user. By standardizing the precipitation of business scenarios and needs, based on the association between standardized scenarios, standardized needs and products, when recommending products, the user portrait under the standardized needs is matched with the product to obtain the product for recommendation to the user, taking into account the different intentions of the product features corresponding to the differentiated needs in different scenarios, and improving the matching accuracy of the recommended products with the user needs, thereby improving the accuracy of product recommendations and the user order rate. In addition, when recommending products to users, based on the standardized precipitation of scenarios and needs, and the association between standardized needs and products, it is convenient for business personnel to focus on user needs without having to understand the full amount of product information in detail in advance, thereby helping business personnel to improve business development efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present application or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 It is a flowchart of the product recommendation method provided in an embodiment of the present application.

[0018] Figure 2 It is a structural diagram of the standardized scenario requirement database provided in an embodiment of the present application.

[0019] Figure 3 It is a schematic diagram of the structure of the standardized product database provided in the embodiment of the present application.

[0020] Figure 4 It is a structural schematic diagram of the product recommendation device provided by the present invention.

[0021] Figure 5 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0023] The embodiment of the present application provides a product recommendation method. In the product recommendation method provided in the embodiment of the present application, standardization is an important means to achieve product recommendation and realize precision marketing. The key implementation path is to standardize and precipitate business scenarios and needs based on the complex needs of users, and then build a product database based on the correspondence between standardized scenarios, standardized needs and products. This path is different from the personalized product recommendation solution based on scenario and demand analysis. Instead, it implements menu-style selection based on the constructed standardized scenarios and standardized needs, and automatically generates a product recommendation pool based on user feature portraits. Through this method, on the one hand, multi-scenario integration and expansion can be achieved, and customer stickiness can be enhanced through multi-scenario product combinations to maximize business expansion benefits; on the other hand, the marketing scenario can be made "simple", which greatly reduces the learning cost of business personnel for business expansion. Business personnel no longer need to understand the characteristics and prices of all products. They only need to check the corresponding standardized scenarios and standardized needs based on the results of the survey on user needs to achieve product recommendations.

[0024] Specifically, refer to Figure 1 , Figure 1 is a flow chart of a product recommendation method provided in an embodiment of the present application, such as Figure 1 As shown, the product recommendation method provided in the embodiment of the present application includes the following steps: Step 100, obtaining standardized requirements corresponding to the target user's purchase intention and historical sales data under the standardized requirements; Step 200, obtaining each product to be recommended under the standardized demand from a standardized product database to obtain a product recommendation pool; the standardized product database is obtained by associating standardized demands and products under various standardized scenarios, and the standardized demands are obtained by performing natural language processing on demand documents under various standardized scenarios; Step 300: construct a user profile under the standardized requirements according to the historical sales data, and match the target product from the product recommendation pool based on the user profile and recommend it to the target user.

[0025] First, obtain the standardized demand corresponding to the target user's purchase intention information and the historical sales data under the standardized demand, wherein the purchase intention can be obtained by business personnel through a survey of user needs, which includes the target user's demand description for the product he intends to purchase, and perform text parsing based on natural language processing on the description information of the intended product contained in the user's purchase intention, and match the text parsing result with the preset standardized demands, so as to determine the standardized demand corresponding to the target user's purchase intention, and the historical sales data is obtained based on the historical purchase records of each user under the standardized demand.

[0026] Optionally, the standardized requirements can also be obtained by the business personnel investigating the purchase intention of the target user and then checking or clicking on the preset standardized requirements in the interactive interface. Optionally, the standardized requirements can also be obtained by the target user checking or clicking on the preset standardized requirements in the interactive interface according to his own purchase intention, which is not specifically limited here.

[0027] The target user can be any user, and different users are distinguished by unique user identifiers. In one embodiment, the target user's purchase intention can be obtained through an interactive interface. For example, after the business personnel conducts a survey on user needs, they configure and input the target user's demand description information for the intended product in the interactive interface by checking, clicking, or manually entering the information, and perform text parsing based on natural language processing on the demand description information input by the business personnel, and match the parsing results with the preset standardized requirements to obtain the standardized requirements of the target user. After obtaining the target user's purchase intention, according to the standardized requirements corresponding to the target user's purchase intention, the historical purchase records of each user for the product under the standardized requirements are obtained to obtain the corresponding historical sales data.

[0028] Determine the standardized needs corresponding to the target user according to the target user's purchase intention, specifically determine the standardized needs of the target user in the standardized scenario. The target user's needs may be compound needs, so the target user's standardized needs may include one or more. The compound needs may correspond to one standardized scenario or a compound scenario. The compound scenario includes multiple standardized scenarios, which can meet the target user's demand selection in multiple scenarios. In one embodiment, after investigating user needs, the business personnel determine the standardized scenario of the target user, and select one or more standardized needs corresponding to the target user's needs in the standardized scenario by checking or clicking, etc., to form the demand information of the target user's purchase intention. Therefore, the target user's standardized needs can be determined based on the purchase intention.

[0029] The products to be recommended under the standardized needs of the target users are obtained from the standardized product database to obtain a product recommendation pool, wherein the standardized product database is obtained by associating the standardized needs and products under each standardized scenario, and the standardized needs are obtained by natural language processing of the demand documents under each standardized scenario. That is, the standardized product database is constructed based on the association between standardized scenarios, standardized needs and products, and the standardized product database contains the association between standardized scenarios and standardized needs, as well as the association between standardized needs and products.

[0030] When obtaining the standardized needs of target users, text parsing is performed on the demand description information of the products that the target users intend to purchase, and the text parsing results are matched with the standardized needs in the standardized product database to obtain the standardized needs corresponding to the target users, and the products associated with the standardized needs are obtained to form a product recommendation pool. Alternatively, the standardized needs under each standardized scenario in the standardized product database are set in the interactive interface for business personnel or users to select by clicking or checking, so as to obtain the standardized needs of target users.

[0031] In the standardized product database, each standardized scenario has at least one standardized requirement, and each standardized requirement is associated with at least one product. If the target user has multiple standardized requirements, the product under each standardized requirement is obtained as the product to be recommended to form a product recommendation pool.

[0032] When recommending products to target users, we build a user profile based on the historical sales data under the standardized needs of the target users. Then, based on the built user profile, we match the target products from the product recommendation pool and recommend them to the target users.

[0033] When constructing a user profile, the sold products under the standardized demand can be determined, and then the purchased products of each user can be obtained, and the user profile of each user can be constructed based on the purchased products of each user. Alternatively, a user profile can be constructed uniformly based on the sold products under the standardized demand.

[0034] In one embodiment, when constructing a user profile for each user, the constructed user profile under the standardized demand includes the user profile of each user who purchases the product under the standardized demand. Specifically, for the sold products under the standardized demand, the purchased products of each user under the standardized demand are determined, the product features of the purchased products of each user are extracted, and the user profile of the user under the standardized demand is constructed based on the product features. When recommending products, based on the user profile of the target user, target products that are the same or similar to the products purchased by the target user are matched from the recommendation pool and recommended to the target user.

[0035] In other embodiments, when constructing a unified user profile for all users who have purchased products under the standardized demand, the product features of the sold products under the standardized demand are extracted, and a unified user profile under the standardized demand is constructed based on the extracted product features of the sold products. When making product recommendations, target products with the same or similar features as the sold products are matched from the product recommendation pool based on the constructed user profile and recommended to the target user.

[0036] Optionally, the product features used to construct the user portrait may be product features that users generally pay attention to, and / or product features that have the greatest impact on the product's order conversion rate, such as price, delivery time, and after-sales service response time.

[0037] When recommending products to target users, we first decompose the target users' purchasing intentions, determine the target users' standardized needs in standardized scenarios, and use the products associated with the target users' standardized needs as the products to be recommended to form a product recommendation pool. Then, we create user profiles based on the products sold under the standardized needs. Finally, we match the target products from the product recommendation pool based on the user profiles and recommend them to the target users.

[0038] In this embodiment, based on the standardized product database, by performing demand decomposition on the user's purchase intention, the user's standardized needs are determined, and the products to be recommended under the standardized needs of the user are obtained from the standardized product database to form a product recommendation pool, and then a user portrait is constructed based on the historical sales data under the standardized needs, and the target product is matched from the product recommendation pool based on the constructed user portrait and recommended to the user. By standardizing the precipitation of business scenarios and needs, based on the association between standardized scenarios, standardized needs and products, when recommending products, the user portrait under the standardized needs is matched with the product to obtain the product for recommendation to the user, taking into account the different intentions of the product features corresponding to the differentiated needs in different scenarios, the matching accuracy of the recommended products and user needs is improved, thereby improving the accuracy of product recommendations and the user order rate. In addition, when recommending products to users, based on the standardized precipitation of scenarios and needs, and the association between standardized needs and products, it is convenient for business personnel to focus on user needs without having to understand the full amount of product information in detail in advance, thereby helping business personnel to improve business development efficiency.

[0039] Before obtaining the standardized needs corresponding to the target user's purchase intention, it is necessary to build a standardized product database. When building a standardized product database, first, by comprehensively sorting out the standardized scenarios involved in the company's business activities, and based on the natural language processing method, processing the demand documents under the standardized scenarios, the needs involved in the standardized scenarios are standardized and divided, and a standardized scenario demand database of "standardized scenarios-standardized needs" is built to achieve rapid retrieval and matching of products based on standardized scenarios and needs, which can not only reduce the time cost of business personnel for demand analysis, but also facilitate the expansion of business with multi-scenario integration.

[0040] Secondly, sort out all product data, including each product's description, price, delivery time, after-sales response time, sales volume, etc., and match products with standardized requirements to form a standardized product database of "standardized scenarios-standardized requirements-products". This can not only greatly reduce the learning pressure of business personnel on product information, but also lay the foundation for product recommendations based on standardized requirements.

[0041] Finally, we select product features that are of general concern to users and have a greater impact on the user's order conversion rate, such as price, product delivery time, and after-sales service response time, and set them as the influencing feature factors of the user's order conversion rate for user portraits. In the product recommendation stage, we use historical sales data to obtain historically purchased products under standardized needs, and perform similarity analysis with the product features of the recommended products to achieve more accurate product recommendations based on refined user portraits.

[0042] In one embodiment, when constructing a standardized product database, firstly, the text of the demand document in each standardized scenario is segmented based on a pre-trained hidden Markov model to obtain a corresponding demand word set, and the TF-IDF (Term Frquency-Inverse Document Frequency) algorithm is used to perform word frequency statistics on each word in the demand word set, and the TF-IDF value of each word is calculated, so as to select the demand keyword with the highest TF-IDF value in the demand document. Based on the pre-trained word2vec algorithm, the vector representation of each demand keyword is obtained, and the demand keywords are clustered based on the vector representation, and the words with the highest frequency are selected from each category obtained by clustering as standardized requirements.

[0043] Furthermore, the description information of each product is segmented based on the pre-trained hidden Markov model to obtain the corresponding product description word set, and then the TF-IDF algorithm is used to perform word frequency statistics on the product word set after the description information segmentation processing to extract product keywords from the product description information.

[0044] Finally, the text similarity between the extracted product keywords and the standardization requirements is calculated, so that the products with similarity greater than the set benchmark value are associated with the standardization requirements, and a standardized product database is constructed that contains the relationship between standardization scenarios, standardization requirements and products.

[0045] Based on this, in step 100, before obtaining the standardized requirements corresponding to the target user's purchase intention, the following may also be included: Step 001, obtaining full product data and demand documents under each preset standardized scenario; the full product data includes product description information of each product; Step 002, reading the demand text in the demand document, and performing word segmentation and stop word removal processing on the demand text and the product description information, respectively, to obtain a first word set corresponding to the demand document and a second word set corresponding to the product description information; Step 003, calculating a first word frequency feature value of each word in the demand text based on the first word set, and calculating a second word frequency feature value of each word in the product description information based on the second word set; Step 004, selecting a plurality of word combinations with the largest first word frequency characteristic value from the demand text as process keyword groups, and selecting a second preset number of words from the product description information as description keywords for the product; Step 005, clustering and grouping the process keyword phrases, and selecting the words with the highest word frequency in each group as the standardized requirements, constructing the corresponding relationship between the standardized requirements and the standardized scenarios, and generating a standardized scenario requirements database; Step 006, performing text similarity analysis on the standardized requirements and the description keywords to determine the text similarity between the standardized requirements and each of the products; Step 007, based on the standardized scenario requirement database, the products whose text similarity is greater than a preset reference value are associated with the standardized requirements to obtain a standardized product database.

[0046] When building a standardized product database, we first obtain the full product data and the demand documents for each preset standardized scenario. The full product data includes the product description information of each product. The full product data can further include the name, price, delivery time, after-sales service response time and sales volume of each product.

[0047] For the acquired demand documents, the text information in each demand document is read to obtain the corresponding demand text, and then the read demand text and the product description information of each product are respectively segmented and stop words are removed to obtain a first word set corresponding to the demand document and a second word set corresponding to the product description information.

[0048] In one embodiment, the demand text of each demand document is segmented and stop words are removed to obtain a corresponding first word set, and the product description information of each product is segmented and stop words are removed to obtain a corresponding second word set.

[0049] The first word frequency feature value of each word in the demand text is calculated based on the first word set, and the second word frequency feature value of each word in the product description information of each product is calculated based on the second word set. In one embodiment, the first word frequency feature value and the second word frequency feature value can be obtained through word frequency statistics.

[0050] Furthermore, based on the calculated first word frequency characteristic value of the demand document and the second word frequency characteristic value of the product description information, a combination of several words with the largest first word frequency characteristic value is selected from the demand text as a process keyword group, and a second preset number of words are selected from the product description information as product description keywords. The process keyword groups are clustered and grouped, and the words with the highest word frequency in each group are selected as standardized requirements, and the corresponding relationship between standardized requirements and standardized scenarios is constructed to generate a standardized scenario requirement database of "standardized scenario-standardized requirements". Figure 2 As shown, Figure 2 Three scenarios, namely, standardized scenario A, standardized scenario B and standardized scenario C, are given as examples, wherein standardized scenario A includes standardized requirements A1-A5, standardized scenario B includes standardized requirements B1-B5, and standardized scenario C includes standardized requirements C1-C5.

[0051] The text similarity analysis is performed on the description keywords of the standardized requirements and the products to determine the text similarity between the description keywords of the standardized requirements and the products. Based on the constructed standardized scenario requirements database, the products with text similarity greater than the preset benchmark value are associated with the standardized requirements to obtain the following results: Figure 3 The standardized product database of “standardized scenarios-standardized requirements-products” shown.

[0052] exist Figure 3 In the figure, it is exemplified that a standardized requirement A1 under a standardized scenario A is associated with m products, such as product A11, product A12, ... product A1m, and a standardized requirement C5 under a standardized scenario C is associated with n products, such as product C51, product C52, ... product C5n.

[0053] It should be noted that the same standardization scenario may correspond to one or more standardization requirements, the number of standardization requirements in different standardization scenarios may be the same or different, and some of the same standardization requirements may exist in different standardization scenarios. Furthermore, the same standardization requirement may be associated with one or more products, and the number of products associated with different standardization requirements may be the same or different. Different standardization requirements in the same standardization scenario may be associated with one or more identical products, and standardization requirements in different standardization scenarios may be associated with one or more identical products.

[0054] Among them, when performing text similarity analysis on the description keywords of standardized requirements and products, specifically, for each standardized requirement, a text similarity analysis is performed on it and the description keywords of each product. Then, based on the set similarity benchmark value, products whose text similarity with the standardized requirement is higher than the set benchmark value are screened out, and associated with the standardized requirement, so as to construct a standardized product database. As for the correspondence between standardized requirements and standardized scenarios, specifically, the requirement documents are classified according to the standardized scenarios, and based on the standardized scenarios to which the requirement documents corresponding to the standardized requirements belong, the standardized scenarios corresponding to the standardized requirements are determined, so as to construct the correspondence between the standardized requirements and the standardized scenarios.

[0055] Furthermore, both the first word frequency feature value and the second word frequency feature value are obtained based on processing of natural language. Specifically, step 003 further includes: Step 0031, counting the first word frequency of the first target word in the first word set in the target document, and the first document quantity of the demand document containing the first target word; the first target word is any word in the first word set, and the target document is the demand document to which the first target word belongs; Step 0032, calculating a first inverse document frequency of the first target word in the demand document according to the total number of documents in the demand document and the first number of documents; Step 0033, calculating the product of the first word frequency and the first inverse document frequency to obtain a first word frequency feature value of the first target word; Step 0034, counting the second word frequency of the second target word in the second word set in the target description information, and the first product quantity of the product containing the second target word in the product description information; the second target word is any word in the second word set, and the target description information is the product description information of the product to which the second target word belongs; Step 0035, calculating a second inverse document frequency of the second target word in the product description information according to the total product quantity of the product and the first product quantity; Step 0036, calculate the product of the second word frequency and the second inverse document frequency to obtain the second word frequency feature value of the second target word.

[0056] When calculating the first frequency feature value of each word in the demand text, firstly, the first frequency of the first target word in the first word set in the target document is counted, and the first document number of the demand document containing the first target word in each demand document is counted. The first target word is any word in the first word set, and the target document is the demand document to which the first target word belongs.

[0057] Optionally, for the first target word, when counting its word frequency in the target document, first count the number of times the first target word appears in the target document, then count the total number of times all words in the target document appear, calculate the ratio of the number of times the first target word appears in the target document to the total number of times all words in the target document appear, and obtain the word frequency of the first target word in the target document.

[0058] According to the total number of documents in the demand documents and the first number of demand documents containing the first target word, the first inverse document frequency of the first target word in all demand documents is calculated, and the product of the first word frequency and the first inverse document frequency is calculated to obtain the first word frequency feature value of the first target word in the target document. In this way, the first word frequency feature value of each word in the first word set is calculated, specifically, the first word frequency feature value of each word in the first word set in the demand document to which it belongs is calculated.

[0059] Similarly, the second frequency feature value of each word in the second word set is calculated in a manner substantially the same as the first target word. Specifically, the second frequency of the second target word in the second word set in the target description information and the first product quantity of the product containing the second target word in the product description information are counted. The second target word is any word in the second word set, and the target description information is the product description information of the product to which the second target word belongs.

[0060] According to the total number of products and the first number of products, the second inverse document frequency of the second target word in the product description information of all products is calculated, and the product of the second term frequency and the second inverse document frequency of the second target word is further calculated to obtain the second term frequency feature value of the second target word in the target description information. In this way, the second term frequency feature value of each word in the second term set is calculated, specifically, the second term frequency feature value of each word in the second term set in the product description information of the product to which it belongs is calculated.

[0061] In one embodiment, the word segmentation processing of the demand text and product description information of the demand document is implemented based on a pre-trained hidden Markov model, and the calculation of the first word frequency feature value and the second word frequency feature value is implemented based on the TF-IDF algorithm. The processing of the demand text and product description information can be performed in parallel or sequentially, and there is no specific limitation on this.

[0062] In one embodiment, according to the business scope of the enterprise, the requirement documents of all projects are sorted out, the requirement documents are matched with the standardized scenarios corresponding to each project, and then the requirement documents are represented as a string set. , Indicates the i-th requirement document. In the word segmentation processing stage, the requirement text of each requirement document is read, and then the pre-trained hidden Markov model is used to perform word segmentation on the requirement text, and the word segmentation result is further processed by removing stop words, etc. to obtain the first word set.

[0063] Optionally, during the pre-training stage of the hidden Markov model, part of the requirement documents can be used to construct a training corpus, and the corpus can be segmented into word labels, including B (beginning of a word), M (middle of a word), E (end of a word), and S (single-word words). During the annotation process of the requirement document corpus, the words describing the requirements are not separated in the middle. For example, "the hotel needs to improve the living experience by installing smart door locks in guest rooms" can be expressed as "BESBESBEBEBMMEBEBEBE". The initial probability matrix, transition probability matrix, and emission matrix of the model are calculated, and then the Viterbi algorithm is used to segment the requirement text of each requirement document to obtain the first word set corresponding to the requirement document. ,in, Represents the jth word in the i-th paragraph of text.

[0064] Optionally, each requirement document may correspond to a first word set, or all requirement documents may correspond to a first word set.

[0065] Based on the first word set, process keywords in the requirement document are extracted. Specifically, based on the TF-IDF algorithm, the requirement document is As the target document, first calculate the word as shown in the following formula 1 In the requirements document The first word frequency in: ; (1)

[0066] In formula 1, It's a word In the requirements document The denominator represents the number of occurrences in the document The sum of the number of occurrences of all words in .

[0067] Then, the word is calculated according to the following formula 2 The first inverse document frequency of the word Inverse document frequency in the document set D corresponding to all requirement documents: ; (2)

[0068] Where N is the total number of requirement documents in the document set D, Indicates that the document set D contains the word The number of requirement documents, that is, the number of first documents.

[0069] Finally, the word frequency is calculated based on the calculated first word frequency and first inverse document frequency. In the documentation The first word frequency feature value in , that is, the first TF-IDF value: ; (3)

[0070] According to the above method, the first TF-IDF value of each word in the first word set is calculated, and according to the calculated first TF-IDF value, several words with the highest first TF-IDF value are selected for each requirement document to form a process keyword group.

[0071] Furthermore, the process keyword groups are clustered and grouped, so as to merge and optimize the process keyword groups, and further extract the words with the highest word frequency as the standardization requirements. Based on this, in step 005, clustering and grouping the process keyword groups specifically includes: Step 0051, extracting the vector representation of each process keyword in the process keyword group; Step 0052, calculating the cosine similarity between each of the process keywords based on the vector representation, and obtaining a similarity matrix corresponding to the process keyword group; Step 0053: Based on the similarity matrix and a preset similarity threshold, an agglomerative hierarchical clustering algorithm is used to cluster and group the process keyword phrases.

[0072] When clustering process keyword groups, the vector representation of each process keyword in the process keyword group is extracted. Based on the extracted vector representation, the cosine similarity between the process keywords is calculated to obtain the similarity matrix of the process keyword group. Based on the similarity matrix and the preset similarity threshold, the agglomerative hierarchical clustering algorithm is used to cluster the process keyword groups.

[0073] In one embodiment, process keywords with the same or similar semantics are merged by using the word2vec algorithm and the agglomerative hierarchical clustering algorithm. Specifically, based on the first word set obtained by the word segmentation of the requirement document, a pre-trained word2vec algorithm model is used to obtain the vector representation of each process keyword. Then, the cosine similarity between the process keywords is calculated to obtain a cosine similarity matrix. For any two vector representations A and B, the representations in the n-dimensional space are respectively and , the calculation method of cosine similarity is shown in the following formula 4: ; (4)

[0074] Based on the preset similarity threshold, the agglomerative hierarchical clustering algorithm is used to group the process keywords, and the words with the highest frequency in each group are selected as the standardization requirements.

[0075] Furthermore, for the extraction of product description keywords, first obtain the product description information of all products, then use the pre-trained hidden Markov model to perform word segmentation on the product description information of each product, and based on the word segmentation result, remove stop words and other processing to obtain the second word set of the product description information. Optionally, the product description information is similar to the requirement document, and the product description information of each product may correspond to a second word set, or the product description information of all products may correspond to a second word set, which will not be repeated here.

[0076] Then, the second word frequency feature value of each word in the second word set is calculated by the TF-IDF algorithm. Specifically, when counting the second word frequency of a word in the product description information, it is specifically based on the product description information of any product, and the number of occurrences of any word and the sum of the number of occurrences of all words are counted. According to the method shown in the above formula 1, the ratio of the two is calculated as the second word frequency of the word in the product description information of the corresponding product. Then, the number of products containing the word in the product description information is counted, and the second inverse document frequency of the word in the product description information of the corresponding product is calculated in the manner shown in the above formula 2 in combination with the total number of products. Finally, the product of the first word frequency and the second inverse document frequency is calculated according to formula 3 to obtain the second TF-IDF value of the word in the product description information of the corresponding product. In this way, the second TF-IDF value of each word in the second word set corresponding to the product description information can be calculated.

[0077] Finally, for each product, a second preset number of keywords with the highest second TF-IDF value are extracted as the product description keywords, and the standardized requirements and the description keywords of each product are subjected to text similarity analysis to obtain the text similarity between the standardized requirements and each product. Based on the set text similarity benchmark value, products with text similarity higher than the benchmark value are screened out, and they are associated with the corresponding standardized requirements, thereby constructing a standardized product database. The text similarity between the standardized requirements and the description keywords of each product can be cosine similarity.

[0078] In some embodiments, as business activities continue to develop, standardized scenarios are adjusted accordingly. Therefore, the scenario design of standardized scenarios can be periodically updated according to actual business conditions, and the above process can be repeated to update standardized requirements and products, thereby updating the standardized product database.

[0079] It is understandable that as products become more abundant, the types of products and related knowledge that front-line business personnel need to learn and understand are gradually increasing. In related technologies, user needs are not associated with products. When there is a lot of product information to learn, it is easy for business personnel to be unable to fully grasp all product information and its corresponding specific needs, and eventually face the dilemma of being unable to accurately recommend products that match their needs to users. In this embodiment, through the word segmentation processing of the hidden Markov model, the word frequency feature calculation of the TF-IDF algorithm, and the text similarity analysis, the standardized needs are matched and associated with the products, and a standardized product database of "standardized scenarios-standardized needs-products" is constructed, which realizes the information call of all products under standardized needs, greatly reduces the learning pressure of business personnel on product information, and can effectively improve the efficiency of product expansion.

[0080] Furthermore, with the development of business activities, the traditional personalized demand decomposition method lacks continuous and effective precipitation and iterative update of user needs due to the uncertainty of the dimensions and scope of the previous demand decomposition, which will increase the learning pressure of business personnel and the time cost of sales orders. In this embodiment, the natural language processing method is adopted, based on the hidden Markov model, TF-IDF algorithm, word2vec algorithm and agglomerative hierarchical clustering algorithm, etc., to learn and analyze the project's demand documents, and finally output standardized requirements based on standardized scenarios in the form of lists, etc., and can achieve periodic iterative optimization of standardized requirements. In terms of algorithm selection, the hidden Markov model is used for word segmentation, and the words describing the demand class can be identified as a complete word during the training process, such as "smart door lock", so that reasonable word segmentation of the demand can be achieved in the process of analyzing the full amount of demand documents. The TF-IDF algorithm is used for keyword extraction, which can filter out common and insignificant words while retaining important words that affect the entire text. The word2vec algorithm and agglomerative hierarchical clustering algorithm are used to merge keywords. The word2vec algorithm can better identify semantics, thereby merging words with the same or similar semantics. Through the above methods, a standardized product database of "standardized scenarios-standardized requirements-products" can be efficiently constructed and iteratively updated, significantly improving the efficiency of business personnel's business development and the quality of demand analysis.

[0081] In the process of business expansion, business personnel often encounter situations where they cannot recommend the best products. On the one hand, there are a large number of different brands of products under the same demand; on the other hand, the same user will have different product feature tendencies under different standardized scenarios and standardized demands. When the relevant product recommendation scheme recommends products based on user portraits, the user portraits constructed are relatively macro and general, and do not take into account the differentiated demand characteristics of users under different demands, which leads to the problem of inaccurate product matching. In this embodiment, based on factors such as price, product delivery time, and after-sales service response time that are generally concerned by users and have a greater impact on the user's order rate, the user portrait of the user under standardized demand is constructed according to the historical sales data under standardized demand. At the same time, the product features corresponding to the influencing feature factors such as price, product delivery time, and after-sales service response time of each product are extracted, and then the recommended products are matched for each standardized demand through similarity analysis. By making user portraits based on standardized demands, the matching accuracy of products and user needs is improved, thereby improving the order rate.

[0082] Based on this, step 300 further includes: Step 301, determining the sold products under the standardized requirements according to the historical sales data; Step 302, based on preset influencing characteristic factors, extract the first product characteristic vector of the sold product, and extract the second product characteristic vector of each product to be recommended in the product recommendation pool; the influencing characteristic factors are product characteristics that have the greatest impact on the user's order conversion rate, and the influencing characteristic factors include product price, product delivery time, and after-sales service response time; Step 303, calculating a user preference vector according to the first product feature vector to construct a user portrait; Step 304, calculating the similarity between the user preference vector corresponding to the user portrait and the second product feature vector of each of the products to be recommended; Step 305: sort the products to be recommended in descending order according to the similarity, and select a first preset number of products to be recommended as target products according to the sorting order and recommend them to the target user.

[0083] When constructing a user portrait and matching target products from the product recommendation pool to recommend to target users based on the user portrait, first determine the sold products under standardized demand based on the historical sales data under standardized demand, extract the first product feature vector of the sold products based on the preset influencing feature factors, and extract the second product feature vector of each product to be recommended in the product recommendation pool. The influencing feature factor is the product feature that has the greatest impact on the user's order conversion rate, and the product features include product price, product delivery time and after-sales service response time.

[0084] Furthermore, the user preference vector is calculated based on the first product feature vector of the sold product, thereby constructing a user portrait. Optionally, if the sold products under the standardized demand include multiple products, the average value of the first product feature vectors of the multiple sold products is calculated to obtain the user preference vector of the user under the standardized demand, and the user preference vector is the user portrait under the standardized demand. If there is only one sold product under the standardized demand, the first product feature vector of the sold product is used as the user preference vector of the user under the standardized demand to obtain the user portrait under the standardized demand. In some embodiments, the purchased products of the target user can be obtained from the sold products according to the user identifier of the target user. If the sold products include the purchased products of the target user, the average value of the first product vectors of the purchased products of the target user is calculated to obtain the user preference vector of the target user under the standardized demand, thereby obtaining the user portrait of the target user under the standardized demand.

[0085] Calculate the similarity between the user preference vector corresponding to the user portrait and the second product feature vector of each product to be recommended. The similarity can also be cosine similarity. Sort the products to be recommended in descending order according to the calculated similarity, so as to select a first preset number of products to be recommended as target products in the sorting order and recommend them to the target user.

[0086] Through the purchase intention of the target user, the standardized demand in the specific scenario is obtained. According to the historical sales data under the standardized demand, including the product name and purchase frequency of the sold products, the historical sales data is associated with the standardized demand based on the constructed standardized product database. Based on the product features that users generally pay attention to and have a greater impact on the customer's order conversion rate, such as product price, product delivery time, and after-sales service response time, they are used as influencing feature factors for user profiling and extracting product feature vectors for each product to be recommended. In the product recommendation stage, according to the user portrait, the target products with the same or similar characteristics as the sold products are matched from the recommended products under the standardized demand and recommended to the target user, so as to recommend products that meet their preferences to the user and improve the accuracy of product recommendations.

[0087] For example, the product feature vector is ,in, Indicates the product price, Indicates the product delivery time. Indicates the after-sales service response time of the product.

[0088] If n products have been sold under a certain standardized demand, the user preference vector U is the average value of the product feature vectors of n products, that is: ; (5)

[0089] Taking standardized demand as the unit, the similarity between the product feature vector of each recommended product associated with the standardized demand of the target user and the user preference vector is calculated by using methods such as cosine similarity: ; (6)

[0090] According to the calculated similarity, the products to be recommended are sorted in descending order from high to low similarity, and then the three target products with the highest similarity are selected in the sorted order and recommended to the target user.

[0091] Furthermore, after step 300, the following steps may be further included: Step 310, obtaining a selection operation for the target product; Step 320: Generate a product solution configuration sheet and a price list based on the target product selected by the selection operation.

[0092] Business personnel select standardized scenarios and standardized requirements in turn as needed, and finally select the product combination and product quantity from the product recommendation pool for each standardized requirement, obtain the selection operation for the target product recommended in the product recommendation pool, and generate a product solution configuration sheet and price list based on the target product selected by the selection operation.

[0093] In one embodiment, due to the lack of understanding of some products by business personnel and the difficulty in product selection, business personnel are unable to quickly and efficiently recommend products and calculate quotations during business development, which has caused great obstacles to business development. Based on this, in the product recommendation process, the embodiment of the present application selects the most suitable products from the product recommendation pool based on the user's scenario and demand information, enters the product quantity, and makes the final product recommendation. After completing the product recommendation, the product recommendation pool will automatically generate a price list of product combination solutions for reference by business personnel, which reduces user waiting time, optimizes the service experience, and effectively improves the order conversion rate.

[0094] In this embodiment, facing the complex user needs, the scenarios and needs are standardized, and a corresponding database is constructed based on the correspondence between the standardized scenarios, standardized needs and products. Then, according to the order conversion rate influencing factors in the selected product features and the historical sales data, user profiling is performed. Among the products to be recommended under the user's standardized needs, the user profiling is further matched with the target products that meet the user's preferences and recommended to the user, thereby improving the quality and efficiency of business development of business personnel.

[0095] Furthermore, the demand documents are divided according to standardized scenarios and processed using natural language processing, including using hidden Markov models for word segmentation, using TF-IDF algorithm to extract keywords, using word2vec and agglomerative hierarchical clustering algorithms to merge keywords, etc., to build a "standardized scenario-standardized demand" database, and can update standardized scenarios and standardized demands according to the business development cycle. In the process of expanding business, business personnel can obtain user demand information through user surveys and other methods, and then check the standardized scenarios according to the "standardized scenario-standardized demand" database, and then check the standardized demands under the standardized scenarios, which can quickly and efficiently generate standardized scenarios and demand information for user needs.

[0096] Furthermore, based on the "standardized scenarios-standardized requirements" database, the same natural language processing method as the requirements document is used to process the product description information and extract the product description keywords, so as to perform text similarity analysis between the product and the standardized requirements based on the product description keywords, and achieve matching association between the product and the standardized requirements. In the process of expanding business, all products under standardized requirements can be automatically obtained, and automatic matching between requirements and products can be achieved, which improves recommendation efficiency and simplifies the operations of business personnel.

[0097] Furthermore, based on the standardized product database of "standardized scenarios-standardized needs-products", the product features that users generally pay attention to and have the greatest impact on the order rate are used as the characteristic factors affecting the order rate. Based on the historical sales data under standardized needs, the user's product feature vector is extracted to create a user portrait, and the user preference vector is obtained. The product feature vector of the product to be recommended is extracted, and the user preference vector and the product feature vector are analyzed for similarity, so as to recommend the target product with the highest similarity to the user preference vector to the user. Based on the correlation between standardized needs and products, and the rapid selection of standardized needs by business personnel, in the process of expanding business, the best product under each standardized need can be automatically analyzed and generated for recommendation, thereby greatly improving the efficiency of business expansion.

[0098] The product recommendation device provided in the embodiment of the present application is described below. The product recommendation device described below and the product recommendation method described above can refer to each other.

[0099] Reference Figure 4 , the product recommendation device provided in the embodiment of the present application includes: The demand acquisition module 10 is used to acquire the standardized demand corresponding to the purchase intention of the target user and the historical sales data under the standardized demand; The product acquisition module 20 is used to acquire each product to be recommended under the standardized demand from a standardized product database to obtain a product recommendation pool; the standardized product database is obtained by associating standardized demands and products under various standardized scenarios, and the standardized demands are obtained by performing natural language processing on demand documents under various standardized scenarios; The product recommendation module 30 is used to construct a user portrait under the standardized requirements according to the historical sales data, and match the target product from the product recommendation pool based on the user portrait and recommend it to the target user.

[0100] In one embodiment, the product recommendation module 30 is further used to: Determine the sold products under the standardized requirements according to the historical sales data; Based on preset influencing characteristic factors, extract the first product characteristic vector of the sold product, and extract the second product characteristic vector of each product to be recommended in the product recommendation pool; the influencing characteristic factors are the product characteristics that have the greatest impact on the user's order conversion rate, and the influencing characteristic factors include product price, product delivery time, and after-sales service response time; Calculating a user preference vector based on the first product feature vector to construct a user portrait; Calculating the similarity between the user preference vector corresponding to the user portrait and the second product feature vector of each of the products to be recommended; The products to be recommended are sorted in descending order according to the similarity, and a first preset number of products to be recommended are selected in the sorting order as target products to be recommended to the target user.

[0101] In one embodiment, the product recommendation device further includes a database construction module, which is used to: Obtaining full product data and demand documents for each preset standardized scenario; the full product data includes product description information of each product; Reading the demand text in the demand document, and performing word segmentation and stop word removal on the demand text and the product description information, respectively, to obtain a first word set corresponding to the demand document and a second word set corresponding to the product description information; Calculate a first word frequency feature value of each word in the demand text based on the first word set, and calculate a second word frequency feature value of each word in the product description information based on the second word set; Selecting a plurality of word combinations with the largest first word frequency characteristic value from the demand text as process keyword groups, and selecting a second preset number of words from the product description information as description keywords; Clustering and grouping the process keyword phrases, and selecting the words with the highest word frequency in each group as the standardized requirements, constructing a corresponding relationship between the standardized requirements and the standardized scenarios, and generating a standardized scenario requirements database; Performing text similarity analysis on the standardized requirements and the description keywords to determine the text similarity between the standardized requirements and each of the products; On the basis of the standardized scenario requirement database, the products whose text similarity is greater than a preset reference value are associated with the standardized requirements to obtain a standardized product database.

[0102] In one embodiment, the database construction module is further used to: Counting the first word frequency of a first target word in the first word set in a target document, and the first document quantity of a demand document containing the first target word; the first target word is any word in the first word set, and the target document is the demand document to which the first target word belongs; Calculating a first inverse document frequency of the first target term in the demand document according to the total number of documents in the demand document and the first number of documents; Calculate the product of the first word frequency and the first inverse document frequency to obtain a first word frequency feature value of the first target word; Counting a second word frequency of a second target word in the second word set in the target description information, and a first product quantity of the product containing the second target word in the product description information; the second target word is any word in the second word set, and the target description information is product description information of the product to which the second target word belongs; Calculating a second inverse document frequency of the second target word in the product description information according to the total product quantity of the product and the first product quantity; The product of the second word frequency and the second inverse document frequency is calculated to obtain a second word frequency feature value of the second target word.

[0103] In one embodiment, the database construction module is further used to: Extracting a vector representation of each process keyword in the process keyword group; Calculate the cosine similarity between each of the process keywords based on the vector representation to obtain a similarity matrix corresponding to the process keyword group; Based on the similarity matrix and a preset similarity threshold, an agglomerative hierarchical clustering algorithm is used to cluster and group the process keyword phrases.

[0104] In one embodiment, the product recommendation device further includes a quotation module, which is used to: Obtaining a selection operation for the target product; A product solution configuration sheet and a price list are generated according to the target product selected by the selection operation.

[0105] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor 510, a communications interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communications interface 520 and the memory 530 communicate with each other via the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute the steps of the product recommendation method, for example including: Obtaining standardized demands corresponding to the target user's purchase intentions and historical sales data under the standardized demands; Acquire each product to be recommended under the standardized demand from a standardized product database to obtain a product recommendation pool; the standardized product database is obtained by associating standardized demands and products under various standardized scenarios, and the standardized demands are obtained by performing natural language processing on demand documents under various standardized scenarios; A user profile under the standardized requirements is constructed according to the historical sales data, and a target product is matched from the product recommendation pool based on the user profile and recommended to the target user.

[0106] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0107] On the other hand, an embodiment of the present application further provides a computer program product, the computer program product including a computer program, the computer program can be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer can perform the steps of the product recommendation method provided by the above methods, for example, including: Obtaining standardized demands corresponding to the target user's purchase intentions and historical sales data under the standardized demands; Acquire each product to be recommended under the standardized demand from a standardized product database to obtain a product recommendation pool; the standardized product database is obtained by associating standardized demands and products under various standardized scenarios, and the standardized demands are obtained by performing natural language processing on demand documents under various standardized scenarios; A user profile under the standardized requirements is constructed according to the historical sales data, and a target product is matched from the product recommendation pool based on the user profile and recommended to the target user.

[0108] On the other hand, the embodiment of the present application further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the product recommendation method provided by the above methods, for example, including: Obtaining standardized demands corresponding to the target user's purchase intentions and historical sales data under the standardized demands; Acquire each product to be recommended under the standardized demand from a standardized product database to obtain a product recommendation pool; the standardized product database is obtained by associating standardized demands and products under various standardized scenarios, and the standardized demands are obtained by performing natural language processing on demand documents under various standardized scenarios; A user profile under the standardized requirements is constructed according to the historical sales data, and a target product is matched from the product recommendation pool based on the user profile and recommended to the target user.

[0109] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0110] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A product recommendation method, characterized in that: include: Obtaining standardized demands corresponding to the target user's purchasing intentions and historical sales data under the standardized demands; Acquire each product to be recommended under the standardized demand from a standardized product database to obtain a product recommendation pool; the standardized product database is obtained by associating standardized demands and products under various standardized scenarios, and the standardized demands are obtained by performing natural language processing on demand documents under various standardized scenarios; A user profile under the standardized requirements is constructed according to the historical sales data, and a target product is matched from the product recommendation pool based on the user profile and recommended to the target user.

2. The product recommendation method according to claim 1, characterized in that: The step of constructing a user profile under the standardized requirements according to the historical sales data, and matching a target product from the product recommendation pool based on the user profile and recommending it to the target user includes: Determine the sold products under the standardized requirements according to the historical sales data; Based on preset influencing characteristic factors, extract the first product characteristic vector of the sold product, and extract the second product characteristic vector of each product to be recommended in the product recommendation pool; the influencing characteristic factors are the product characteristics that have the greatest impact on the user's order conversion rate, and the influencing characteristic factors include product price, product delivery time, and after-sales service response time; Calculating a user preference vector based on the first product feature vector to construct a user portrait; Calculating the similarity between the user preference vector corresponding to the user portrait and the second product feature vector of each of the products to be recommended; The products to be recommended are sorted in descending order according to the similarity, and a first preset number of products to be recommended are selected in the sorting order as target products to be recommended to the target user.

3. The product recommendation method according to claim 1, characterized in that: Before obtaining the standardized requirements corresponding to the target user's purchase intention, the method further includes: Obtaining full product data and demand documents for each preset standardized scenario; the full product data includes product description information of each product; Reading the demand text in the demand document, and performing word segmentation and stop word removal on the demand text and the product description information, respectively, to obtain a first word set corresponding to the demand document and a second word set corresponding to the product description information; Calculate a first word frequency feature value of each word in the demand text based on the first word set, and calculate a second word frequency feature value of each word in the product description information based on the second word set; Selecting a plurality of word combinations with the largest first word frequency characteristic value from the demand text as process keyword groups, and selecting a second preset number of words from the product description information as description keywords; Clustering and grouping the process keyword phrases, and selecting the words with the highest word frequency in each group as the standardized requirements, constructing a corresponding relationship between the standardized requirements and the standardized scenarios, and generating a standardized scenario requirements database; Performing text similarity analysis on the standardized requirements and the description keywords to determine the text similarity between the standardized requirements and each of the products; On the basis of the standardized scenario requirement database, the products whose text similarity is greater than a preset reference value are associated with the standardized requirements to obtain a standardized product database.

4. The product recommendation method according to claim 3, characterized in that: The calculating a first word frequency feature value of each word in the demand text based on the first word set, and calculating a second word frequency feature value of each word in the product description information based on the second word set, includes: Counting the first word frequency of a first target word in the first word set in a target document, and the first document quantity of a demand document containing the first target word; the first target word is any word in the first word set, and the target document is the demand document to which the first target word belongs; Calculating a first inverse document frequency of the first target term in the demand document according to the total number of documents in the demand document and the first number of documents; Calculate the product of the first word frequency and the first inverse document frequency to obtain a first word frequency feature value of the first target word; Counting a second word frequency of a second target word in the second word set in the target description information, and a first product quantity of the product containing the second target word in the product description information; the second target word is any word in the second word set, and the target description information is product description information of the product to which the second target word belongs; Calculating a second inverse document frequency of the second target word in the product description information according to the total product quantity of the product and the first product quantity; The product of the second word frequency and the second inverse document frequency is calculated to obtain a second word frequency feature value of the second target word.

5. The product recommendation method according to claim 3, characterized in that: The clustering and grouping of the process keyword phrases comprises: Extracting a vector representation of each process keyword in the process keyword group; Calculate the cosine similarity between each of the process keywords based on the vector representation to obtain a similarity matrix corresponding to the process keyword group; Based on the similarity matrix and a preset similarity threshold, an agglomerative hierarchical clustering algorithm is used to cluster and group the process keyword phrases.

6. The product recommendation method according to claim 1, characterized in that: After matching the target product from the product recommendation pool based on the user portrait and recommending it to the target user, the method further includes: Obtaining a selection operation for the target product; A product solution configuration sheet and a price list are generated according to the target product selected by the selection operation.

7. A product recommendation device, characterized in that: include: A demand acquisition module is used to acquire the standardized demand corresponding to the purchase intention of the target user and the historical sales data under the standardized demand; A product acquisition module is used to obtain each product to be recommended under the standardized demand from a standardized product database to obtain a product recommendation pool; the standardized product database is obtained by associating standardized demands and products under various standardized scenarios, and the standardized demands are obtained by performing natural language processing on demand documents under various standardized scenarios; The product recommendation module is used to construct a user portrait under the standardized requirements according to the historical sales data, and match the target product from the product recommendation pool based on the user portrait and recommend it to the target user.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the product recommendation method according to any one of claims 1 to 6 are implemented.

9. A non-transitory 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 product recommendation method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the product recommendation method according to any one of claims 1 to 6 are implemented.

Images