Product recommendation method, device, electronic device and storage medium

Abstract

The embodiment of the present application discloses a product recommendation method, device, electronic device and storage medium, the method comprising: after obtaining the candidate product names of candidate merchants according to the user's request, grouping the candidate product names to obtain multiple product groups, and determining the product group names; sorting the product group names based on the product group name sorting model according to the user's historical behavior data as the recommendation order of the multiple product groups; assigning a product display position to each product group according to the recommendation order, and displaying a product in the corresponding product group in the product display position as the first-level display result of the multiple product groups; recommending the first-level display result to the user. Since the embodiment of the present application can recommend a product group instead of a merchant when recommending, it can facilitate users to compare different products, shorten the user's decision path, reduce the user's decision time, and improve user experience.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a product recommendation method, device, electronic device, and storage medium. Background Art

[0002] Existing food delivery recommendation systems simply recommend businesses to users. They primarily estimate the probability of users clicking on or placing an order for a displayed business based on their historical behavior and preferences, and then rank businesses based on probability.

[0003] Existing takeout recommendation systems create a lengthy decision-making process for users: first, users must select a merchant, then choose a dish, calculate the price, and potentially even compare prices before placing an order. This lengthy decision-making process prevents users from quickly selecting their favorite, high-quality dishes, leading to lengthy decision-making times and negative user experience and conversion rates. Summary of the Invention

[0004] The embodiments of the present application provide a product recommendation method, device, electronic device, and storage medium to shorten the user's decision-making path and reduce the user's decision-making time.

[0005] To solve the above problems, in a first aspect, embodiments of the present application provide a product recommendation method, comprising:

[0006] After obtaining candidate product names of candidate merchants according to the user's request, grouping the candidate product names to obtain multiple product groups and determining product group names;

[0007] sorting the product group names based on a product group name sorting model according to the user's historical behavior data as a recommended order for the multiple product groups;

[0008] Allocate a product display position to each product group according to the recommendation order, and display one product from the corresponding product group in the product display position as a primary display result of the multiple product groups;

[0009] The first-level display result is recommended to the user.

[0010] In a second aspect, an embodiment of the present application provides a product recommendation device, comprising:

[0011] A product grouping module is used to, after obtaining candidate product names of candidate merchants according to a user's request, group the candidate product names to obtain multiple product groups and determine the product group names;

[0012] a product group sorting module, configured to sort the product group names based on the user's historical behavior data and a product group name sorting model, as a recommended order for the multiple product groups;

[0013] a first-level display determination module, configured to assign a product display position to each product group according to the recommendation order, and display one product from the corresponding product group in the product display position as a first-level display result for the multiple product groups;

[0014] A recommendation module is used to recommend the first-level display result to the user.

[0015] In a third aspect, an embodiment of the present application further discloses an electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the product recommendation method described in the embodiment of the present application when executing the computer program.

[0016] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the product recommendation method disclosed in the embodiment of the present application are performed.

[0017] The product recommendation method, device, electronic device and storage medium disclosed in the embodiments of the present application obtain candidate product names of candidate merchants according to a user's request, group the candidate product names to obtain multiple product groups, and determine the product group names. According to the user's historical behavior data, the product group names are sorted based on a product group name sorting model as the recommendation order of the multiple product groups. According to the recommendation order, a product display position is assigned to each product group, and a product in the corresponding product group is displayed at the product display position as a first-level display result of the multiple product groups. The first-level display result is recommended to the user. Since a product group can be recommended instead of a merchant, it is convenient for the user to compare different products, shortening the user's decision-making path, reducing the user's decision-making time, and improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0019] Figure 1 This is a flowchart of the product recommendation method of Example 1 of the present application;

[0020] Figure 2This is a flowchart of determining the first-level display result in an embodiment of the present application;

[0021] Figure 3 This is a page diagram of the first-level display results in the embodiment of the present application;

[0022] Figure 4 This is a page diagram of the secondary display results in the embodiment of the present application;

[0023] Figure 5 It is a structural diagram of the product recommendation device of Example 2 of the present application. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described 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 making creative efforts are within the scope of protection of this application.

[0025] Example 1

[0026] This embodiment discloses a product recommendation method, such as Figure 1 As shown, the method includes: steps 110 to 140.

[0027] Step 110 : After obtaining the candidate product names of the candidate merchants according to the user's request, the candidate product names are grouped to obtain a plurality of product groups, and product group names are determined.

[0028] The candidate merchants can be determined based on the user's geographic location, such as merchants within a 1000-meter radius of the user's geographic location. The candidate merchants can be takeout merchants. The product group name is a combination of the product name and the product category. For example, if the product name is Mango Salmon Roll and the product category is Cold Dishes, the product group name can be Salmon.

[0029] After receiving the user's request, candidate merchants can be determined based on the user's geographic location, and all products in each candidate merchant can be used as candidate products to obtain candidate product names. When grouping the candidate product names, the candidate product names can be clustered based on the product attributes of each candidate product name, thereby dividing the candidate product names into multiple product groups. The same characteristics or attributes of the products in a product group can be used as the product group name; alternatively, the product names can be clustered offline to obtain a mapping relationship between product names and product group names. Based on the predetermined mapping relationship between product names and product group names, the product group name corresponding to each candidate product name is determined, and the candidate product names with the same product group name are grouped into one product group, and the product group name is used as the product group name corresponding to the product group.

[0030] In one embodiment of the present application, grouping the candidate product names to obtain multiple product groups and determining the product group names includes: determining the product group names corresponding to the candidate product names based on the candidate product names and the mapping relationship between the product names and the product group names, thereby obtaining the multiple product groups. By predetermining the mapping relationship between the product names and the product group names, the candidate product names can be grouped accurately and quickly, thereby improving processing speed and, in turn, improving recommendation speed.

[0031] In one embodiment of the present application, before determining the product group name corresponding to the candidate product name based on the candidate product name and the mapping relationship between the product name and the product group name, it also includes: extracting the main words of the product name as the primary standard name corresponding to the product name; extracting the product attributes corresponding to the primary standard name based on a sequence labeling algorithm; determining the hierarchical relationship of the primary standard name based on the product attributes corresponding to the primary standard name; determining the product group name corresponding to the hierarchical relationship based on the hierarchical relationship; mapping the product name to the product group name based on the primary standard name corresponding to the product name and the hierarchical relationship to obtain the mapping relationship between the product name and the product group name.

[0032] Taking the candidate merchant as a takeaway merchant and the candidate product as a takeaway product as an example, the product attributes may include category, ingredients, taste and cooking method.

[0033] When extracting the main words of a product name, an HMM (Hidden Markov Model) algorithm can be used. For example, for a product named "Spicy Chicken Stewed Rice," the main word extracted could be "Spicy Chicken Stewed Rice," which would then serve as the primary standard name for the product name "Spicy Chicken Stewed Rice." An LSTM (Long Short-Term Memory) sequence tagging algorithm can be used to extract the product attributes corresponding to each primary standard name. For example, for the primary standard name "Spicy Chicken Stewed Rice," the ingredients are chicken, the cooking method is "braised rice," and the flavor is "spicy pepper." The ingredients, cooking method, and flavor are then used as product attributes. You can first determine the same product attributes of each primary standard name, determine that there is a primary standard name with the same attribute in the product attributes that belongs to a hierarchical relationship, and then determine the hierarchical relationship of each primary standard name based on the hierarchical relationship between the product attributes of each primary standard name in the same hierarchical relationship. For example, the primary standard names are salmon, salmon roll, salmon fruit roll, and mango salmon roll. The ingredients of the primary standard name "Salmon" are salmon, the ingredients of the primary standard name "Salmon Roll" are salmon and seasoning (such as cream cheese), and the ingredients of the primary standard name "Salmon Fruit Roll" are salmon. Fish, fruit and seasonings, the ingredients of the primary standard name "Mango Salmon Roll" are salmon, mango and seasonings, and mango is a kind of fruit. Therefore, "Mango Salmon Roll" is at the bottom layer, and its upper layer is "Salmon Fruit Roll". Compared with "Salmon Fruit Roll", "Salmon Roll" has one less ingredient, so "Salmon Roll" is located in the upper layer of "Salmon Fruit Roll". Compared with "Salmon Roll", "Salmon" has one less ingredient, so "Salmon" is located in the upper layer of "Salmon Roll", that is, "Salmon" is located in the top layer. At this point, the hierarchical relationship of the above primary standard names is determined. After the hierarchical relationship is determined, for each hierarchical relationship, the primary standard name at the top layer is used as the product group name corresponding to the hierarchical relationship. Then, based on the primary standard names belonging to the same hierarchical relationship and the product names corresponding to the primary standard names, the product names are mapped to the product group names corresponding to the hierarchical relationship to obtain the mapping relationship between product names and product group names.

[0034] Step 120 : sorting the product group names based on the product group name sorting model according to the user's historical behavior data as a recommended order for the multiple product groups.

[0035] The product group name ranking model is a pre-trained machine learning model. This machine learning model can be a neural network model, a decision tree model, or other machine learning model. The goal of the product group name ranking model is to predict the probability of a user placing an order for products within a product group name, sort the product group names based on the order probability, and determine the recommended order for multiple product groups. Historical user behavior data includes user exposure data and order data.

[0036] Before training the product group name ranking model, first construct the product group model training data: obtain all the product names that have not been sold within a preset time (such as a week), and map the product name to the corresponding product group name, and the data corresponding to the product group name is used as a negative sample; obtain all the product names that have been sold within a preset time (such as a week), map the product name to the corresponding product group name, and the data corresponding to the product group name is used as a positive sample; the positive and negative samples both include user identification, product group identification and context features, extract user features, product group name features and user-to-product group name interaction features, and combine user features, product group name features and user-to-product group name interaction features based on the positive and negative samples to obtain product group model training data. After the product group model training data is constructed, the machine learning model is trained, and after the training is completed, the product group name ranking model is obtained, and the product group name ranking model is synchronized online for use in online product group name sorting. Among them, user features include user gender, age, whether white-collar, income level, consumption level, average consumption price, etc. Product group name features include the price of the product group name (the average price of all products in the product group name), the 7-day / 30-day / 90-day sales volume of the product group name (the average sales volume of all products in the product group name), the review score of the product group name (the average review score of all products in the product group name), the CTR (Click-Through-Rate) of the unique visitors or visits to the product group name, the CVR (Conversion Rate) of the unique visitors or visits to the product group name, and the order conversion rate of the unique visitors or visits to the product group name. CTR, CVR, and order conversion rate are all calculated by mapping the product name to the corresponding product group name: CTR = clicks / impressions, CVR = orders / clicks, and order conversion rate = orders / impressions. User interaction features for product group names include the 7-day / 30-day / 90-day click count, order count, CTR, CVR, and order conversion rate. Each feature is calculated by mapping the product name to the corresponding product group name and then calculating the user's relevant features for the product group name. Contextual features include city, current time period, and client type.

[0037] Step 130 : Allocate a product display position to each product group according to the recommendation order, and display a product in the corresponding product group at the product display position as a primary display result of the multiple product groups.

[0038] Each product group is assigned a product display slot in the recommended order. This slot displays a product from the corresponding product group, allowing multiple product groups to be displayed in the recommended order. The product to be displayed in the product display slot can be randomly selected from the product group, or the candidate product names within the product group can be sorted and the first-ranked candidate product name can be selected as the product to be displayed in the product display slot.

[0039] In one embodiment of the present application, Figure 2 As shown, according to the recommendation order, a product display position is allocated to each product group, and one product from the corresponding product group is displayed at the product display position as a primary display result of the multiple product groups, including steps 131 to 133:

[0040] Step 131 : sorting the candidate product names corresponding to each product group based on the product sorting model according to the user's historical behavior data, and obtaining a product sorting result corresponding to each product group.

[0041] The product ranking model is a pre-trained machine learning model. It can be a neural network model, a decision tree model, or other machine learning model. The goal of the product ranking model is to predict the probability of a user placing an order for an exposed product, sort the products based on the order probability, and determine the order in which the products are exposed. Historical user behavior data includes both user exposure data and order data.

[0042] Before training the product ranking model, the product model training data is first constructed: obtain the product data of all exposed products that have not been sold within a preset time (such as a week), and use these product data as negative samples. Obtain the product data of all exposed products that have been sold within a preset time (such as a week), and use these product data as positive samples. Both the positive and negative samples include user identification, product identification, and context features. User features are extracted based on the user identification, product features are extracted based on the product identification, and user-to-product interaction features are extracted based on the user identification and product identification. The user features, product features, and user-to-product interaction features are combined to obtain product model training data. After the product model training data is constructed, the machine learning model is trained based on the product model training data. After the training is completed, a product ranking model is obtained, and the product ranking model is synchronized online for use in online product ranking. Among them, user features include user gender, age, whether white-collar worker, income level, consumption level, and average consumption price. Product features include price, 7-day / 30-day / 90-day sales volume, review score, CTR of unique visitors or page views, CVR of unique visitors or page views, and conversion rate of unique visitors or page views. User-to-product interaction features include CTR, CVR, and conversion rate for the product, as well as CTR, CVR, and conversion rate for the product's category. Contextual features include city, current time period, and client type.

[0043] In a specific embodiment, the candidate product names corresponding to each product group are sorted based on the product sorting model according to the user's historical behavior data to obtain the product sorting results corresponding to each product group, including: for each product group, extracting user characteristics, candidate product name characteristics and user-candidate product name interaction characteristics as first input characteristics according to the user's historical behavior data; processing the first input characteristics corresponding to the product group through the product sorting model to obtain the user's order probability for each candidate product name in the product group as the first order probability; sorting the candidate product names corresponding to the product group in descending order of the first order probability to obtain the product sorting results corresponding to each product group. For each product group, the candidate products under the product group are sorted. Based on the user's historical behavior data, user characteristics, candidate product name characteristics and user-candidate product name interaction characteristics are extracted. These characteristics are used as the first input features to input into the product sorting model. The product sorting model predicts the user's order probability for each candidate product name as the first order probability. The candidate product names corresponding to a product group are sorted in descending order of the first order probability. The sorting result is used as the product sorting result corresponding to the product group. The products under the product group can be displayed according to the product sorting result.

[0044] It should be noted that step 131 and step 120 can be executed in parallel, thereby increasing the processing speed and further increasing the recommendation speed.

[0045] Step 132 : Determine the preferred display products corresponding to each product group according to the product ranking results corresponding to each product group.

[0046] After obtaining the product ranking result corresponding to each product group, the product ranked first in each product group is used as the preferred display product corresponding to the product group.

[0047] Step 133 : Allocate a product display position for each product group according to the recommendation order, and display the preferred display products corresponding to each product group in the product display position of the corresponding product group, thereby obtaining the primary display results of the multiple product groups.

[0048] like Figure 3 As shown, each product group corresponds to a product display position, which is used to display the preferred display products of the product group. Figure 3 Taking a first-level display result display page showing four product groups as an example, product group 1, product group 2, product group 3, and product group 4 can be displayed. Each product group occupies a product display position and displays the preferred display product in the product group.

[0049] Step 140: recommend the primary display result to the user.

[0050] The first-level display results are recommended to users. Multiple product groups can be displayed on the user side. The products in each product group have the same characteristics, which makes it easy for users to browse different product groups and select products after determining the product group.

[0051] In one embodiment of the present application, the product recommendation method also includes: using the product sorting results corresponding to each product group as the secondary display results of the corresponding product group; if the user's operation instructions for the specified preferred display product are received through the page of the first-level display results, the secondary display results corresponding to the product group to which the specified preferred display product belongs are sent to the user.

[0052] The product ranking result corresponding to a product group is the secondary display result of the product group, that is, the products under the product group are displayed in the order of the product ranking. If the user clicks on the preferred display product of a product group on the page of the primary display result on the user side, the server will send the secondary display result corresponding to the product group to which the designated preferred display product belongs to the user side, and the user side will display the products under the product group to which the designated preferred display product belongs in the order of the product ranking in the secondary display result. Figure 4 As shown, on the secondary result display page at the user end, the products under a product group are displayed according to the product sorting results, and each product occupies a product display position. Figure 4 For example, displaying 4 products, that is, product 1, product 2, product 3 and product 4 can be displayed in order. When the user slides up and down, the display page can display other products under the product group according to the product sorting results.

[0053] In one embodiment of the present application, the method of sorting the product group names based on the product group name sorting model according to the user's historical behavior data as a recommended order for the multiple product groups includes: extracting user characteristics, product group name characteristics, and user-product group name interaction characteristics based on the user's historical behavior data as second input characteristics; processing the second input characteristics through the product group name sorting model to obtain the user's order probability for each product group name as the second order probability; and sorting the product group names in descending order of the second order probabilities as the recommended order for the multiple product groups. When sorting the product group names, the method of first extracting user characteristics, product group name characteristics, and user-product group name interaction characteristics based on the user's historical behavior data as the second input characteristics, processing the second input characteristics through the product group name sorting model, predicting the user's order probability for the product group name, that is, predicting the user's order probability for the products in the product group as the second order probability, and sorting the product group names in descending order of the second order probabilities to obtain a recommended order for the multiple product groups.

[0054] The product recommendation method disclosed in the embodiments of the present application obtains candidate product names from candidate merchants in response to a user's request, then groups the candidate product names to obtain multiple product groups and determines product group names. The product group names are then sorted based on a product group name ranking model based on the user's historical behavior data, and the ranking is used as a recommendation order for the multiple product groups. A product display position is assigned to each product group according to the recommendation order, and one product from the corresponding product group is displayed in the product display position as a primary display result for the multiple product groups. This primary display result is then recommended to the user. Since a product group can be recommended instead of a merchant, it facilitates user comparison of different products, shortens the user's decision-making process, reduces decision-making time, and improves the user experience. Furthermore, for the recommended merchants, the recommended product group can expose more merchants, thereby increasing their exposure rate.

[0055] Example 2

[0056] This embodiment discloses a product recommendation device, such as Figure 5 As shown, the product recommendation device 500 includes:

[0057] The product grouping module 510 is configured to, after obtaining candidate product names of candidate merchants according to a user's request, group the candidate product names to obtain multiple product groups and determine product group names;

[0058] A product group ranking module 520 is configured to sort the product group names based on the user's historical behavior data and a product group name ranking model to serve as a recommended order for the multiple product groups;

[0059] A primary display determination module 530 is configured to assign a product display position to each product group according to the recommendation order, and display a product from the corresponding product group in the product display position as a primary display result for the multiple product groups;

[0060] The recommendation module 540 is configured to recommend the primary display result to the user.

[0061] Optionally, the first-level display determination module includes:

[0062] a product ranking unit, configured to sort the candidate product names corresponding to each product group based on the product ranking model according to the user's historical behavior data, and obtain a product ranking result corresponding to each product group;

[0063] A preferred product determination unit, configured to determine the preferred display product corresponding to each product group based on the product sorting result corresponding to each product group;

[0064] The display determination unit is used to allocate a product display position to each product group according to the recommendation order, and display the preferred display products corresponding to each product group in the product display position of the corresponding product group to obtain the primary display results of the multiple product groups.

[0065] Optionally, the device further includes:

[0066] A secondary display determination module is used to use the product ranking result corresponding to each product group as the secondary display result of the corresponding product group;

[0067] The secondary display sending module is used to send the secondary display result corresponding to the product group to which the specified first-choice display product belongs to the user if an operation instruction for specifying the first-choice display product is received from the user through the page of the primary display result.

[0068] Optionally, the product sorting unit includes:

[0069] a feature extraction subunit, configured to extract, for each product group, user features, candidate product name features, and user-to-candidate product name interaction features based on the user's historical behavior data, as first input features;

[0070] an order probability prediction subunit, configured to process the first input feature corresponding to the product group using the product ranking model to obtain a probability of the user placing an order for each candidate product name in the product group as a first order probability;

[0071] The product sorting subunit is configured to sort the candidate product names corresponding to the product groups in descending order of the first order probability, to obtain a product sorting result corresponding to each product group.

[0072] Optionally, the product grouping module includes:

[0073] The product grouping unit is configured to determine the product group name corresponding to the candidate product name according to the candidate product name and the mapping relationship between the product name and the product group name, thereby obtaining a plurality of product groups.

[0074] Optionally, the device further includes:

[0075] A primary standard name determination module is used to extract the main words of the product name as the primary standard name corresponding to the product name;

[0076] A product attribute extraction module, configured to extract the product attributes corresponding to the primary standard name based on a sequence labeling algorithm;

[0077] A hierarchical relationship determination module, configured to determine the hierarchical relationship of the primary standard names according to the commodity attributes corresponding to the primary standard names;

[0078] A commodity group name determination module, configured to determine, based on the hierarchical relationship, the commodity group name corresponding to the hierarchical relationship;

[0079] A mapping relationship determination module is used to map the product name to the product group name according to the primary standard name corresponding to the product name and the hierarchical relationship, so as to obtain a mapping relationship between the product name and the product group name.

[0080] Optionally, the product group sorting module includes:

[0081] a feature extraction unit, configured to extract, based on the user's historical behavior data, user features, product group name features, and user-to-product group name interaction features as second input features;

[0082] an order probability prediction unit, configured to process the second input feature using the product group name ranking model to obtain a probability of the user placing an order for each product group name as a second order probability;

[0083] The recommendation order determining unit is configured to sort the product group names in descending order of the second order probability as a recommendation order for the plurality of product groups.

[0084] The product recommendation device provided in the embodiment of the present application is used to implement the various steps of the product recommendation method described in the embodiment of the present application. The specific implementation methods of each module of the device can be found in the corresponding steps and will not be repeated here.

[0085] The product recommendation device disclosed in the embodiment of the present application, after obtaining the candidate product names of candidate merchants according to the user's request through the product grouping module, groups the candidate product names to obtain multiple product groups and determines the product group names. The product group sorting module sorts the product group names based on the product group name sorting model according to the user's historical behavior data as the recommendation order of multiple product groups. The first-level display determination module allocates a product display position to each product group according to the recommendation order, and displays a product in the corresponding product group at the product display position as the first-level display result of multiple product groups. The recommendation module recommends the first-level display result to the user. Since a product group can be recommended instead of a merchant, it is convenient for the user to compare different products, shortens the user's decision path, reduces the user's decision-making time, and improves the user experience.

[0086] Accordingly, embodiments of the present application further disclose an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the product recommendation method described in the embodiments of the present application is implemented. The electronic device may be a PC, a mobile terminal, a personal digital assistant, a tablet computer, or the like.

[0087] The embodiment of the present application also discloses a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the product recommendation method as described in the embodiment of the present application are implemented.

[0088] Each embodiment in this specification is described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between the various embodiments can be referred to in conjunction with each other. For the device embodiments, since they are generally similar to the method embodiments, their description is relatively simple, and for relevant parts, reference can be made to the description of the method embodiments.

[0089] It should be noted that, in this document, relational terms such as first and second, etc. are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0090] The above is a detailed introduction to a product recommendation method, device, electronic device and storage medium provided in the embodiments of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, based on the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

[0091] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment 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 essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling 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 certain parts of the embodiments.

Claims

1. A product recommendation method, characterized in that: include: After obtaining candidate product names of candidate merchants according to the user's request, grouping the candidate product names to obtain multiple product groups and determining product group names; sorting the product group names based on a product group name sorting model according to the user's historical behavior data as a recommended order for the multiple product groups; Allocate a product display position to each product group according to the recommendation order, and display one product from the corresponding product group in the product display position as a primary display result of the multiple product groups; recommending the primary display result to the user; The step of allocating a product display position to each product group according to the recommendation order, and displaying a product from the corresponding product group at the product display position as a primary display result of the multiple product groups includes: According to the user's historical behavior data, the candidate product names corresponding to each product group are sorted based on the product sorting model to obtain a product sorting result corresponding to each product group; According to the ranking results of the products corresponding to each product group, determine the preferred display products corresponding to each product group; Allocate a product display position for each product group according to the recommendation order, and display the preferred display products corresponding to each product group in the product display position of the corresponding product group, thereby obtaining the primary display results of the multiple product groups; The method further comprises: The product sorting results corresponding to each product group are used as the secondary display results of the corresponding product group; If an operation instruction for specifying a preferred display product is received from the user through the page of the first-level display results, the second-level display results corresponding to the product group to which the specified preferred display product belongs are sent to the user; The step of grouping the candidate product names to obtain a plurality of product groups and determining product group names includes: Determining the product group names corresponding to the candidate product names based on the candidate product names and the mapping relationship between product names and product group names, to obtain multiple product groups; Before determining the product group name corresponding to the candidate product name based on the candidate product name and the mapping relationship between the product name and the product group name, the method further includes: For the product name, extract the main words of the product name as the primary standard name corresponding to the product name; Extracting the commodity attributes corresponding to the primary standard name based on a sequence labeling algorithm; Determining the hierarchical relationship of the primary standard names according to the commodity attributes corresponding to the primary standard names; Determine, based on the hierarchical relationship, a commodity group name corresponding to the hierarchical relationship; According to the primary standard name corresponding to the product name and the hierarchical relationship, the product name is mapped to the product group name to obtain a mapping relationship between the product name and the product group name.

2. The method according to claim 1, characterized in that The method of sorting the candidate product names corresponding to each product group based on the product sorting model according to the user's historical behavior data to obtain a product sorting result corresponding to each product group includes: For each product group, extracting user features, candidate product name features, and user-to-candidate product name interaction features as first input features based on the user's historical behavior data; Processing the first input feature corresponding to the product group using the product ranking model to obtain a probability of the user placing an order for each candidate product name in the product group as a first order probability; The candidate product names corresponding to the product groups are sorted in descending order of the first order probability to obtain a product sorting result corresponding to each product group.

3. The method according to claim 1, characterized in that The step of sorting the product group names based on a product group name sorting model according to the user's historical behavior data as a recommended order for the multiple product groups includes: Extracting user features, product group name features, and user-to-product group name interaction features as second input features based on the user's historical behavior data; Processing the second input feature using the product group name ranking model to obtain a probability of the user placing an order for each product group name as a second order probability; The product group names are sorted in descending order of the second order probability as a recommended order for the multiple product groups.

4. A product recommendation device, characterized in that: include: A product grouping module is used to, after obtaining candidate product names of candidate merchants according to a user's request, group the candidate product names to obtain multiple product groups and determine the product group names; a product group sorting module, configured to sort the product group names based on the user's historical behavior data and a product group name sorting model, as a recommended order for the multiple product groups; a first-level display determination module, configured to assign a product display position to each product group according to the recommendation order, and display one product from the corresponding product group in the product display position as a first-level display result for the multiple product groups; A recommendation module, configured to recommend the primary display result to the user; The first-level display determination module further includes: The step of allocating a product display position to each product group according to the recommendation order, and displaying a product from the corresponding product group at the product display position as a primary display result of the multiple product groups includes: According to the user's historical behavior data, the candidate product names corresponding to each product group are sorted based on the product sorting model to obtain a product sorting result corresponding to each product group; According to the ranking results of the products corresponding to each product group, determine the preferred display products corresponding to each product group; Allocate a product display position for each product group according to the recommendation order, and display the preferred display products corresponding to each product group in the product display position of the corresponding product group, thereby obtaining the primary display results of the multiple product groups; The first-level display determination module further includes: The product sorting results corresponding to each product group are used as the secondary display results of the corresponding product group; If an operation instruction for specifying a preferred display product is received from the user through the page of the first-level display results, the second-level display results corresponding to the product group to which the specified preferred display product belongs are sent to the user; The product grouping module also includes: The step of grouping the candidate product names to obtain a plurality of product groups and determining product group names includes: Determining the product group names corresponding to the candidate product names based on the candidate product names and the mapping relationship between product names and product group names, to obtain multiple product groups; Before determining the product group name corresponding to the candidate product name based on the candidate product name and the mapping relationship between the product name and the product group name, the method further includes: For the product name, extract the main words of the product name as the primary standard name corresponding to the product name; Extracting the commodity attributes corresponding to the primary standard name based on a sequence labeling algorithm; Determining the hierarchical relationship of the primary standard names according to the commodity attributes corresponding to the primary standard names; Determine, based on the hierarchical relationship, a commodity group name corresponding to the hierarchical relationship; According to the primary standard name corresponding to the product name and the hierarchical relationship, the product name is mapped to the product group name to obtain a mapping relationship between the product name and the product group name.

5. 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 product recommendation method according to any one of claims 1 to 3 is implemented.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the product recommendation method according to any one of claims 1 to 3 are implemented.

Images