Commodity recommendation method and device, equipment and medium

Abstract

The invention relates to a commodity recommendation method and device, equipment and medium in the technical field of e-commerce, and the method comprises the steps: responding to a commodity recommendation event of an independent station shop, and determining a target recommendation scene and a target user corresponding to the event, obtaining candidate commodity sets recalled by a plurality of recall sources corresponding to the scene and an initial recall score of each candidate commodity in the candidate commodity sets, and correspondingly adjusting the initial recall score of each candidate commodity in each candidate commodity set to determine a target recall score according to the corresponding effect weight of each recall source compared with other recall sources; when a dynamically generated historical recommendation effect result of an independent station shop represents that user behavior evaluation needs to be carried out on the shop, determining a user behavior score of a target user associated with each recalled candidate commodity, and determining a recommendation score based on a target recall score of each candidate commodity and the user behavior score; and a plurality of recommended commodities are selected and pushed to the target user. According to the invention, the personalized recommendation of the shop is realized.

Description

Technical Field

[0001] The present application relates to the field of e-commerce technology, and in particular to a product recommendation method and its corresponding device, computer equipment, and computer-readable storage medium. Background Art

[0002] Independent websites usually have their own domain names and online stores built on them. Stores have a high degree of autonomy and independence in operation. Each store has its own types, and there are significant differences in terms of operating cycle, business categories, target markets, traffic channels, etc. From the perspective of operating cycle, some stores have been in operation for many years and have more data, while some stores have just been established and have less data in terms of quantity and / or types; from the perspective of business categories, some stores mainly sell clothing products, while others mainly sell 3C products; from the perspective of target markets, some stores have markets in the United States, while others have markets in the United Kingdom; from the perspective of traffic channels, some stores have traffic from Facebook ads, some from Google ads, and some from SEO, etc. In this regard, it is necessary to customize recommendation strategies based on the store's own unique user groups and diverse recall strategies.

[0003] In traditional technology, the product recommendations of integrated platforms usually serve multiple online stores built on the same platform. The recommendation methods adopted for this need to be applicable to the vast majority of online stores, such as recommendations based on collaborative filtering algorithms, recommendations based on user portraits, and recommendations based on vector representations. However, these recommendation methods are difficult to apply in the personalized product recommendations of independent stores. For example, a newly established store may lack sufficient relevant data to support the recommendation effect of collaborative filtering algorithms; stores that mainly sell clothing products may have very different user interests and purchasing behaviors from stores that mainly sell 3C products, and so on.

[0004] In view of the shortcomings of traditional practices, the applicant has been engaged in research in related fields for a long time and has taken a different approach to solve industry problems in the field of e-commerce technology. Summary of the invention

[0005] The primary purpose of the present application is to solve at least one of the above problems and to provide a product recommendation method and its corresponding device, computer equipment, and computer program product.

[0006] In order to meet the various objectives of this application, this application adopts the following technical solutions:

[0007] A product recommendation method provided for one of the purposes of this application comprises the following steps:

[0008] Respond to the product recommendation event of the independent station store and determine the target recommendation scenario and target user corresponding to the event;

[0009] Obtaining candidate product sets respectively recalled by multiple recall sources corresponding to the target recommendation scenario and the initial recall score of each candidate product therein, and adjusting the initial recall score of each candidate product in each candidate product set according to the effectiveness weight corresponding to each recall source compared with other recall sources to determine the target recall score;

[0010] Obtaining dynamically generated historical recommendation results of the independent website store, and when the historical recommendation results indicate that a user behavior evaluation of the store is required, determining a user behavior score of the target user associated with each recalled candidate product;

[0011] A recommendation score is determined according to the target recall score and the user behavior score of each candidate product, and a plurality of recommended products are selected based on the recommendation score sorting and pushed to the target user.

[0012] On the other hand, a product recommendation device provided to meet one of the purposes of the present application includes an event response module, a product recall module, a product evaluation module and a product push module, wherein the event response module is used to respond to a product recommendation event of an independent station store and determine a target recommendation scenario and a target user corresponding to the event; the product recall module is used to obtain a set of candidate products respectively recalled by multiple recall sources corresponding to the target recommendation scenario and an initial recall score of each candidate product therein, and adjust the initial recall score of each candidate product in each candidate product set according to the effectiveness weight corresponding to each recall source compared with other recall sources to determine a target recall score; the product evaluation module is used to obtain the dynamically generated historical recommendation effectiveness results of the independent station store, and when the historical recommendation effectiveness results indicate that a user behavior evaluation needs to be performed on the store, determine the user behavior score of the target user associated with each recalled candidate product; the product push module is used to determine a recommendation score according to the target recall score and the user behavior score of each candidate product, and select multiple recommended products based on the recommendation score sorting and push them to the target user.

[0013] On the other hand, a computer device provided to meet one of the purposes of the present application includes a central processing unit and a memory, wherein the central processing unit is used to call and run a computer program stored in the memory to execute the steps of the product recommendation method described in the present application.

[0014] On the other hand, a computer program product provided to meet another purpose of the present application includes a computer program / instruction, which, when executed by a processor, implements the steps of the method described in any embodiment of the present application.

[0015] The technical solution of this application has many advantages, including but not limited to the following aspects:

[0016] This application first defines the target recommendation scenario and target user corresponding to the product recommendation event of the independent station store, and obtains the candidate product set recalled by the corresponding multiple recall sources and the initial recall score of each candidate product therein for the scenario, and then adjusts the initial recall score of each candidate product in each candidate product set according to the corresponding effectiveness weight of each recall source compared with other recall sources to determine the target recall score. The recall value of the candidate products in the target recommendation scenario can be accurately quantified, so that the recall value of the products with higher recommendation effectiveness in the actual service recommendation scenario is relatively strengthened, and the recall value of the products with higher recommendation effectiveness in the recommendation scenario is weakened. On the basis of achieving a balance between high-value recall and diverse recall, it helps to improve the personalized recommendation effectiveness of the store.

[0017] Secondly, based on the dynamically generated historical recommendation results of the independent station store, when its representation requires user behavior evaluation of the store, the user behavior score of the target user associated with each recalled candidate product is determined, and the association value between the candidate product and the user can be quantified on demand. It can also dynamically update the historical recommendation results accordingly to adapt to the continuous operation of the independent station store, ensuring the timeliness and accuracy of the results, so as to more accurately serve the personalized product recommendations of the independent station store.

[0018] On the above basis, based on the target recall score and user behavior score of each candidate product, the recommendation score is determined, and multiple recommended products are selected and pushed to the target users. By comprehensively evaluating the recall value of the candidate products in the target recommendation scenario and the association value with the user, the recommendation value is accurately determined, so that products that meet the expectations of the target users can be recommended to them for the target recommendation scenario, attracting the user's attention, which is expected to facilitate the final product transaction and achieve high-quality recommendation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0020] Figure 1 The network architecture of the e-commerce platform exemplified in this application;

[0021] Figure 2 A flowchart of a typical embodiment of the product recommendation method of the present application;

[0022] Figure 3 This is a functional block diagram of the product recommendation device of this application;

[0023] Figure 4 A schematic diagram of the structure of a computer device used in this application. DETAILED DESCRIPTION

[0024] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be interpreted as limiting the present application.

[0025] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0026] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with those in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as here.

[0027] like Figure 1 In the network architecture shown, the e-commerce platform 82 is deployed on the Internet to provide corresponding services to its users. Similarly, the devices 80 of the merchant users of the e-commerce platform 82 and the devices 81 of the consumer users are also connected to the Internet to use the services provided by the e-commerce platform.

[0028] The exemplary e-commerce platform 82 provides supply and demand matching of products and / or services to the general public with the help of Internet infrastructure. In the e-commerce platform 82, products and / or services are provided as commodity information. To simplify the description, the concepts of commodity, product, etc. are used in this application to refer to the products and / or services in the e-commerce platform 82, which may specifically be physical products, digital products, tickets, service subscriptions, other offline services, etc.

[0029] In reality, all entities can access the e-commerce platform 82 as users, use various online services provided by the e-commerce platform 82, and achieve the purpose of participating in the business activities achieved by the e-commerce platform 82. These entities can be natural persons, legal persons or social organizations, etc. Corresponding to the two types of entities, merchants and consumers in business activities, the e-commerce platform 82 has two types of users, merchant users and consumer users. All entities in the product distribution chain in business activities, including manufacturers, sellers, retailers, logistics providers, etc., can use online services in the e-commerce platform 82 as merchant users, while consumers in business activities, including real or potential consumers, can use online services in the e-commerce platform 82 as their corresponding consumer users. In actual business activities, the same entity can act as both a merchant user and a consumer user, and this should be understood flexibly.

[0030] The infrastructure used to deploy the e-commerce platform 82 mainly includes the backend architecture and frontend equipment. The backend architecture runs various online services through the service cluster, including middleware or frontend services for the platform, services for consumers, services for merchants, etc., to enrich and improve its service functions; the frontend equipment mainly covers the terminal devices used by users to access the e-commerce platform 82 as clients, including but not limited to various mobile terminals, personal computers, point-of-sale devices, etc. For example, merchant users can enter product information for their online stores through their terminal devices 80, or generate their product information using the interfaces opened by the e-commerce platform; consumer users can access the webpage of the online store implemented by the e-commerce platform 82 through their terminal devices 81, trigger the shopping process through the shopping buttons provided on the webpage, and call various online services provided by the e-commerce platform 82 during the shopping process, so as to achieve the purpose of shopping orders.

[0031] In some embodiments, the e-commerce platform 82 may be implemented by a processing facility including a processor and a memory, the processing facility storing a set of instructions that, when executed, cause the e-commerce platform 82 to perform the e-commerce and support functions involved in the present application. The processing facility may be part of a server, client, network infrastructure, mobile computing platform, cloud computing platform, fixed computing platform, or other computing platform, and provides electronic components, merchant equipment, payment gateways, application developers, marketing channels, transportation providers, customer equipment, point-of-sale equipment, etc. of the e-commerce platform 82.

[0032] The e-commerce platform 82 can be implemented as online services such as cloud computing services, software as a service (SaaS), infrastructure as a service (IaaS), platform as a service (PaaS), desktop as a service (DaaS), hosted software as a service, mobile backend as a service (MBaaS), information technology management as a service (ITMaaS), etc. In some embodiments, the various functional components of the e-commerce platform 82 can be implemented to be suitable for operation on various platforms and operating systems. For example, for an online store, its administrator users enjoy the same or similar functions regardless of various embodiments such as iOS, Android, HomonyOS, or web pages.

[0033] The e-commerce platform 82 can realize the corresponding independent station for each merchant to run its corresponding online store, and provide the merchant with the corresponding business management engine instance for the merchant to establish, maintain and run one or more online stores in one or more independent stations. The business management engine instance can be used for content management, task automation and data management of one or more online stores, and various specific business processes of the online store can be configured through interfaces or built-in components to support the implementation of business activities. The independent station is the infrastructure of the e-commerce platform 82 with cross-border service functions. Merchants can maintain their online stores more centrally and autonomously based on the independent station. The independent station usually has a domain name and storage space dedicated to the merchant, and different independent stations are relatively independent. The e-commerce platform 82 can provide standardized or personalized technical support for a large number of independent stations, so that merchant users can customize their own business management engine instance and use this business management engine instance to maintain one or more online stores they own.

[0034] The online store can implement backend configuration and maintenance by having the merchant user log in to its business management engine instance as an administrator. With the support of various online services provided by the infrastructure of the e-commerce platform 82, the merchant user can configure various functions in its online store as an administrator, view various data, etc. For example, the merchant user can manage various aspects of its online store, such as viewing the online store's recent activities, updating the online store's product catalog, managing orders, recent visit activities, total order activities, etc.; the merchant user can also view more detailed information about the business and visitors to the merchant's online store by obtaining reports or metrics, such as showing a sales summary of the merchant's overall business, specific sales and participation data of active sales marketing channels, etc.

[0035] The e-commerce platform 82 may provide communication facilities and associated merchant interfaces for providing electronic communications and marketing, such as utilizing electronic message aggregation facilities to collect and analyze communication interactions between merchants, consumers, merchant devices, customer devices, point-of-sale devices, etc., aggregating and analyzing communications, such as for increasing the potential for providing product sales, etc. For example, a consumer may have a question related to a product, which may generate a dialogue between the consumer and the merchant (or an automated processor-based agent on behalf of the merchant), where the communication facility is responsible for interacting and providing the merchant with analysis on how to increase the probability of a sale.

[0036] In some embodiments, an application suitable for installation in a terminal device can be provided to serve the access needs of different users, so that various users can access the e-commerce platform 82 by running the application in the terminal device, such as the merchant backend module of the online store in the e-commerce platform 82, etc. In the process of implementing business activities through these functions, the e-commerce platform 82 can implement various functions related to supporting the implementation of business activities as middleware or online services and open corresponding interfaces, and then implant the toolkit corresponding to the interface access function into the application to implement function expansion and task implementation. The business management engine can include a series of basic functions, and expose these functions to online services and / or application calls through APIs. Online services and applications use corresponding functions by remotely calling corresponding APIs.

[0037] With the support of various components of the business management engine instance, the e-commerce platform 82 can provide online shopping functions, allowing merchants to establish connections with customers in a flexible and transparent manner. Consumer users can select items online, create product orders, provide the delivery address of the goods in the product order, and complete the payment confirmation of the product order. Then, the merchant can review and complete or cancel the order. The review component carried by the business management engine instance can realize the compliance use of business processes to ensure that the order is suitable for fulfillment before actual fulfillment. Orders may sometimes be fraudulent and need to be verified (such as ID card checks). There is a payment method that requires merchants to wait to ensure that funds are received, which can play a role in preventing such risks, etc. Order risks may be generated by fraud detection tools submitted by third parties through order risk APIs, etc. Before fulfillment, merchants may need to obtain payment information or wait to receive payment information in order to mark the order as paid, so that the merchant can prepare to deliver the product. Such situations can be reviewed accordingly. The review process can be implemented by the fulfillment component. Merchants can use the fulfillment component to review, adjust work, and trigger related fulfillment services, such as: manual fulfillment service, which is used when merchants pick and pack products in boxes, purchase shipping labels and enter their tracking numbers, or simply mark items as fulfilled; custom fulfillment service, which can define sending emails for notification; API fulfillment service, which can trigger third-party applications to create fulfillment records in third parties; legacy fulfillment service, which can trigger custom API calls from the business management engine to third parties; gift card fulfillment service. It can provide the generation of numbers and activate gift cards. Merchants can use the order printer application to print shipping orders. The fulfillment process can be executed when the items are packed in boxes and ready for shipping, tracking, delivery, consumer receipt verification, etc.

[0038] It can be seen that the services provided by the e-commerce platform are based on products. The corresponding commodity data is the basic data of the e-commerce platform. The commodity information is provided through the commodity data. The mining and utilization of commodity data is the basis for realizing various technical services, including the use of user transaction data and commodity data in the commodity data of the e-commerce platform to provide basic services for the operation of the data processing system. Therefore, the data processing system can run on any one or more servers of the e-commerce platform's cluster, so as to realize various functions by using the various commodity data provided by the e-commerce platform.

[0039] A product recommendation method of the present application can be programmed as a computer program product and deployed in a client or server for execution. For example, in the exemplary application scenario of the present application, it can be deployed and implemented in a server of an e-commerce customer service platform, thereby accessing an interface opened after the computer program product is running and performing human-computer interaction with the process of the computer program product through a graphical user interface to execute the method.

[0040] See also Figure 2 The product recommendation method of the present application, in its typical embodiment, includes the following steps:

[0041] Step S1100, responding to the product recommendation event of the independent website store, and determining the target recommendation scenario and target user corresponding to the event;

[0042] On the one hand, any of the various actions taken by users in the independent station store, such as browsing, searching, clicking, adding to the shopping cart, purchasing, visiting, etc., can trigger product recommendation events. On the other hand, any of the specific pages visited by users in the independent station store, such as the home page, order settlement page, order placement page, shopping cart page, search page, browsing history page, etc., can trigger product recommendation events.

[0043] In an independent website store, for various different product recommendation events that occur therein, recommendation scenarios are pre-configured. The recommendation scenarios may refer to the recommendation of pages and / or behaviors that trigger the corresponding product recommendation events, or may refer to the recommendation of activities formulated based on the operation and / or marketing conditions of the independent website store. In addition, an event identifier is pre-set for each product recommendation event. The event identifier is used to uniquely refer to the corresponding product recommendation event to distinguish it from other product recommendation events, and can be flexibly implemented by those skilled in the art.

[0044] When a user in an independent station store triggers a product recommendation event, the user's terminal device carrying the independent station store constructs a product recommendation request carrying the event identifier of the product recommendation event and sends it to the server.

[0045] The server receives the product recommendation request, responds to the event, takes the user as the target user, and determines the recommendation scenario pre-configured for the event according to the unique identifier as the target recommendation scenario.

[0046] Step S1200: obtaining a set of candidate products respectively recalled by multiple recall sources corresponding to the target recommendation scenario and an initial recall score of each candidate product therein, and adjusting the initial recall score of each candidate product in each candidate product set according to the effectiveness weight corresponding to each recall source compared with other recall sources to determine a target recall score;

[0047] In independent station stores, at least one recall source is usually pre-configured for various recommendation scenarios of its services. Secondly, the same recall source can be applicable to multiple recommendation scenarios. The recall source is applicable to the corresponding recommendation scenario and is implemented based on a recall strategy and / or recall channel. For example, for recommendations whose recommendation scenario is the user's click behavior, since the user clicks on the product because of interest, in order to achieve recommendations that meet the user's expectations, the corresponding configuration of various recall sources may be similar product title recall, similar product category recall, similar product image recall, similar product information recall, etc., to recall products similar to the product clicked by the user; for recommendations whose recommendation scenario is the homepage, in order to show users the products that they may be interested in, the corresponding configuration of various recall sources may be user behavior recall, collaborative filtering recall, store high click recall, store high add-on recall, store high order recall, user region recall, etc., to recall products that users are interested in.

[0048] Each recall source recalls all products from the independent station store, usually quantitatively expressing the initial recall score corresponding to each product, and then selects multiple products with relatively high initial recall scores as candidate products to construct a candidate product set corresponding to the recall source. The initial recall score represents the value of the product being recalled by the corresponding recall source, and the value refers to the degree of match with the recall target of the recall source. It is not difficult to understand that different recall sources have their own recall targets and their respective recall targets are different. For example, the recall target of the similar product title recall is to recall products that are similar to the product title of the specified product. In this regard, the value of the product depends on the similarity between the product title of the product and the product title of the specified product. The quantitative representation of the similarity is the initial recall score.

[0049] To facilitate understanding of the recall source recall in the recommendation scenario, for example, taking the recommendation scenario as the recommendation of the user's click behavior, taking the recall source as the recall of similar product titles as an example, the text encoding model is used to vectorize the product titles of all products in the independent station store, and the text encoding vector corresponding to each product title is obtained. The vector similarity algorithm is used to calculate the similarity between the text encoding vector corresponding to the product title of the clicked product and the text encoding vector of the product title of other products, that is, the initial recall score, and these products are sorted in order from high to low according to the initial recall score, and multiple products with high rankings are selected as candidate products for recall, and all candidate products recalled are collected to form a candidate product set. The text encoding model is pre-trained and learns the ability to encode the input text to obtain the text encoding vector corresponding to the semantics of the text. The selection of the model can be flexibly selected and implemented by technicians in this field. The vector similarity algorithm can be any one of cosine similarity, Jaccard coefficient, Euclidean distance, etc., and technicians in this field can flexibly select and implement it.

[0050] The effectiveness weight quantifies the effectiveness of a single recall source applicable to the same recommendation scenario compared to other recall sources. This can be achieved by first sorting and filtering out the corresponding recommended products for each recall source based on all the products recalled by the recall source and their initial recall scores, and then determining the ratio between the total number of recommended products corresponding to the target user behavior of the user and the total number of recommended products corresponding to the target user behavior of the user among these recommended products, as the effectiveness score of the recall source. Further, the effectiveness scores of each recall source are normalized to obtain the effectiveness weight corresponding to each recall source. The target user behavior can be any one or more of click, add to cart, purchase, and repeat purchase.

[0051] For each candidate product set recalled by a recall source, the initial recall score of each candidate product in the candidate product set is multiplied by the effectiveness weight of the recall source corresponding to the candidate product set to obtain a target recall score corresponding to each candidate product.

[0052] Step S1300: Obtain the dynamically generated historical recommendation results of the independent website store, and when the historical recommendation results indicate that a user behavior evaluation of the store is required, determine the user behavior score of the target user associated with each recalled candidate product;

[0053] When the historical recommendation results show that the recommendation results of the independent station store for user behavior evaluation are better, it indicates that the store needs to be evaluated for user behavior at this time to relatively improve the recommendation results. Therefore, the user feature information of the target user and the product feature information of each candidate product are obtained, and the user behavior score of each candidate product is determined based on the user feature information and each product feature information as input using a pre-trained behavior prediction model. The behavior prediction model learns through training the ability to reliably and accurately predict the user behavior score including any one or more of the click-through rate, conversion rate, and click-to-conversion rate based on the input user feature information and product feature information.

[0054] Step S1400: determine a recommendation score according to the target recall score and the user behavior score of each candidate product, and select a plurality of recommended products based on the recommendation score sorting and push them to the target user.

[0055] Furthermore, for each candidate product, the target recall score and user behavior score of the candidate product are multiplied by their respective weights and then added together to obtain a recommendation score. All recalled candidate products are sorted in descending order of recommendation scores, and multiple candidate products with the highest rankings are selected as recommended products. Then, the recommendation display page of the target recommendation scenario is obtained, and each recommended product is placed on the recommendation display page in sequence and sent to the target user. The number of candidate products to be screened, the weight of the target recall score, and the weight of the user behavior score can all be pre-set by technicians in this field according to business needs.

[0056] It is not difficult to understand from the above embodiments that, compared with the prior art, the present application has many advantages, including at least:

[0057] This application first defines the target recommendation scenario and target user corresponding to the product recommendation event of the independent station store, and obtains the candidate product set recalled by the corresponding multiple recall sources and the initial recall score of each candidate product therein for the scenario, and then adjusts the initial recall score of each candidate product in each candidate product set according to the corresponding effectiveness weight of each recall source compared with other recall sources to determine the target recall score. The recall value of the candidate products in the target recommendation scenario can be accurately quantified, so that the recall value of the products with higher recommendation effectiveness in the actual service recommendation scenario is relatively strengthened, and the recall value of the products with higher recommendation effectiveness in the recommendation scenario is weakened. On the basis of achieving a balance between high-value recall and diverse recall, it helps to improve the personalized recommendation effectiveness of the store.

[0058] Secondly, based on the dynamically generated historical recommendation results of the independent station store, when its representation requires user behavior evaluation of the store, the user behavior score of the target user associated with each recalled candidate product is determined, and the association value between the candidate product and the user can be quantified on demand. It can also dynamically update the historical recommendation results accordingly to adapt to the continuous operation of the independent station store, ensuring the timeliness and accuracy of the results, so as to more accurately serve the personalized product recommendations of the independent station store.

[0059] On the above basis, based on the target recall score and user behavior score of each candidate product, the recommendation score is determined, and multiple recommended products are selected and pushed to the target users. By comprehensively evaluating the recall value of the candidate products in the target recommendation scenario and the association value with the user, the recommendation value is accurately determined, so that products that meet the expectations of the target users can be recommended to them for the target recommendation scenario, attracting the user's attention, which is expected to facilitate the final product transaction and achieve high-quality recommendation results.

[0060] In a further embodiment, after step S1300, obtaining the dynamically generated historical recommendation performance results of the independent website store, the following steps are included:

[0061] Step S2300: When the historical recommendation effectiveness result indicates that user behavior evaluation of the store is not required, the target recall score of each recalled candidate product is used as the recommendation score;

[0062] When the historical recommendation effectiveness result shows that the recommendation effect is better when the user behavior evaluation is not performed on the independent site store, it means that there is no need to perform user behavior evaluation on the store at this time to relatively improve the recommendation effect. Therefore, the target recall score of each recalled candidate product is used as the recommendation score corresponding to the recommendation value of the candidate product.

[0063] Step S2310: Filter out a plurality of recommended products whose recommendation scores meet preset conditions and push them to the target user.

[0064] The recalled multiple candidate products are sorted in descending order of recommendation scores, and then the top N candidate products are selected as recommended products, and these recommended products are collected to form a recommended product list. The N can be set as needed by those skilled in the art.

[0065] Obtain a recommendation display page for the target recommendation scenario, and place the recommended products in the product ranking list in order on the page, that is, the higher the ranking of the recommended products, the closer they are to the top of the page, so that users can browse products with higher recommended scores first based on their regular top-down browsing habits.

[0066] In this embodiment, when there is no need to conduct user behavior evaluation on the independent station store, the target recall score of each recalled candidate product is used as the recommendation score, and then multiple products recommended to the user are selected. This can adapt to the best recommendation needs of the independent station store and recommend products accordingly, thereby achieving the best recommendation results or high-quality recommendation effects as much as possible.

[0067] In a further embodiment, before step S1100 responding to the product recommendation event of the independent website store, starting the performance weight generation process includes the following steps:

[0068] Step S1000: for the multiple preset recommendation scenarios corresponding to the independent website store, determine the multiple recall sources associated with each recommendation scenario;

[0069] In an independent website store, at least one recall source is usually pre-configured for various recommendation scenarios, thereby obtaining at least one recall source configured for each recommendation scenario.

[0070] Step S1010: based on each recall source in each recommendation scenario, recall corresponding target products for the user's product recommendation event, and push the target products recalled by each recall source to the user;

[0071] For each recommendation scenario, respond to the product recommendation event of the user in the recommendation scenario, call each recall source in the recommendation scenario for each product recommendation event, and for each recall source, determine the initial recall score of each product from all the products in the independent station store, sort them in order from high to low according to the initial recall score, select the top-ranked products as target products, and then recall these target products. After that, send these recalled target products to the user who triggered the product recommendation event.

[0072] Step S1020: Obtain effectiveness data generated by exposing the target product to corresponding users, and for each recommendation scenario, determine the effectiveness weight corresponding to each recall source based on the effectiveness data of the target product recalled by each recall source in the recommendation scenario.

[0073] After each target product is exposed to the corresponding user, monitor whether the user performs the target user behavior on the target product. When the user performs the target user behavior on the target product, the performance data of the target product is considered to be met; when the user does not perform the target user behavior on the target product, the performance data of the target product is considered to be unsatisfactory.

[0074] For each recommendation scenario, according to the effectiveness data of the target products recalled by each recall source in the recommendation scenario, the ratio of the total number of target products that meet the effectiveness standards divided by the total number of target products that do not meet the effectiveness standards among all the target products recalled by each recall source is determined as the effectiveness score of the recall source. Further, the obtained effectiveness scores of each recall source are normalized to obtain the effectiveness weight of each recall source in the recommendation scenario. The normalization operation can be implemented by using any normalization algorithm such as Sigma function, z-score standardization, linear scale normalization (Min-Max Sca ling), truncation, binning, logarithmic transformation, etc. In one embodiment of the normalization, the recall sources are sorted in order of effectiveness scores from high to low, and the ranking of each recall source is determined. The reciprocal of the ranking of each recall source is taken as the effectiveness weight of the recall source. For the sake of understanding the exemplary example, the ranking of the recall source is 10, and the effectiveness weight of the recall source is 1 / 10.

[0075] In this embodiment, the effectiveness weights of each recall source associated with each recommendation scenario involved in the independent station store are determined. In this process, on the one hand, the effectiveness weights of each recall source associated with all recommendation scenarios involved in the independent station store are quantified for product recommendation, so that the recommendation accuracy is higher; on the other hand, by considering the recall effectiveness performance of each recall source under the same recommendation scenario, the effectiveness weight of each recall source that is different from other recall sources in the recommendation scenario is determined, which can ensure the accuracy and reliability of the effectiveness weight, thereby more effectively utilizing those recall sources that perform better in actual applications, while reducing the negative impact of those recall sources with poor results on the recommendation results, so that the recommendation effect is better.

[0076] In a further embodiment, step S1300, obtaining the dynamically generated historical recommendation performance results of the independent website store, includes the following steps:

[0077] Step S1310: In response to the timed triggered online recommendation effectiveness measurement event, the online user recommendation flow associated with the same recommendation scenario is divided into the first user recommendation flow and the second user recommendation flow;

[0078] The timing trigger may be caused by the server starting a timing task, which is used to trigger an online recommendation effectiveness measurement event every predetermined time. The task goal is to determine whether the independent station store needs to conduct user behavior evaluation based on whether the user behavior evaluation of the independent station store has improved the recommendation effectiveness. The predetermined time period may be half a month, one month, half a year, one year, etc., and can be preset by technicians in this field according to business needs.

[0079] To ensure reliability and accuracy, scheduled tasks are executed in the online environment of product recommendations in independent station stores. The product recommendation events of all users who are online associated with the same recommendation scenario, that is, user recommendation traffic, are divided into the first user recommendation traffic and the second user recommendation traffic, that is, each contains the same number of product recommendation events.

[0080] Step S1320: for the product recommendation events of each user included in the first user recommendation flow, the target products of each corresponding recall source and their target recall scores are recalled, and after the recommended products are sorted and screened out based on the target recall scores and pushed to the user, the effectiveness data generated by the exposure of the recommended products to the corresponding user is obtained, and the first recommendation effectiveness is determined based on all the effectiveness data;

[0081] For each product recommendation event, determine its corresponding target recommendation scenario and target user, obtain the candidate product set recalled by each recall source associated with the target recommendation scenario, and the initial recall score of each candidate product in each candidate product set, take each candidate set as the target product set, take each candidate product as the target product, multiply the effectiveness weight of the recall source corresponding to each target product set by the initial recall score of each target product in the target product set to obtain the target recall score as the recommendation score. Sort all recalled target products in order of recommendation score from high to low, filter out multiple target products with top ranking as recommended products, then obtain the recommendation display page of the target recommendation scenario, place each recommended product on the recommendation display page in sequence and send it to the target user to complete the response to the product recommendation event. The number of target products to be screened can be pre-set by technicians in this field according to business needs.

[0082] Furthermore, after completing the response to each product recommendation event, it is determined whether the user has performed the target user behavior on the recommended product after each recommended product is sent to the corresponding user. When the user performs the target user behavior on the recommended product, the effectiveness data of the recommended product is generated as the effectiveness meets the standard; when the user does not perform the target user behavior on the recommended product, the effectiveness data of the recommended product is generated as the effectiveness does not meet the standard. Afterwards, the ratio of the total number of recommended products that meet the effectiveness standard divided by the total number of recommended products that do not meet the effectiveness standard among all recommended products is determined as the first recommendation effectiveness.

[0083] Step S1330: for the product recommendation events of each user included in the second user recommendation flow, the target products of each corresponding recall source and their target recall scores are recalled, and the user behavior score of the corresponding user associated with each target product is determined. After the recommended products are sorted and screened based on the target recall score and the user behavior score and pushed to the user, the effectiveness data generated by the exposure of the recommended products to the corresponding user is obtained, and the second recommendation effectiveness is determined based on all the effectiveness data;

[0084] For each product recommendation event, determine its corresponding target recommendation scenario and target user, obtain the candidate product set recalled by each recall source associated with the target recommendation scenario, and the initial recall score of each candidate product in each candidate product set, take each candidate set as the target product set, take each candidate product as the target product, multiply the effectiveness weight of the recall source corresponding to each target product set by the initial recall score of each target product in the target product set to obtain the target recall score. In addition, obtain the user feature information of the target user and the product feature information of each target product, use the pre-trained to convergence behavior prediction model to use the user feature information and each product feature information as input, and determine the user behavior score of each target product. Further, for each target product, multiply the target recall score and user behavior score of the target product by their respective weights and then add them together to obtain the recommendation score. All recalled target products are sorted in descending order of recommendation scores, and multiple target products with the highest ranking are selected as recommended products. Then, a recommendation display page of the target recommendation scenario is obtained, and each recommended product is placed on the recommendation display page in sequence and sent to the target user to complete the response to the product recommendation event. The number of target products to be screened, the weight of the target recall score, and the weight of the user behavior score can all be pre-set by those skilled in the art according to business needs.

[0085] Furthermore, after completing the response to each product recommendation event, it is determined whether the user has performed the target user behavior for each recommended product after the recommended product is sent to the corresponding user. When the user performs the target user behavior for the recommended product, the effectiveness data of the recommended product is generated as the effectiveness meets the standard; when the user does not perform the target user behavior for the recommended product, the effectiveness data of the recommended product is generated as the effectiveness does not meet the standard. Afterwards, the ratio of the total number of recommended products that meet the effectiveness standard divided by the total number of recommended products that do not meet the effectiveness standard among all recommended products is determined as the second recommendation effectiveness.

[0086] Step S1340: When the first recommendation effect is higher than the second recommendation effect, confirm that the historical recommendation effect result is that the recommendation effect of performing user behavior evaluation on the independent station store is better; otherwise, confirm that the historical recommendation effect result is that the recommendation effect of not performing user behavior evaluation on the independent station store is better.

[0087] When the first recommendation effect is higher than the second recommendation effect, the result shows that conducting user behavior evaluation on the independent station store improves the recommendation effect. Therefore, it is confirmed that the historical recommendation effect result is that conducting user behavior evaluation on the independent station store has better recommendation effect.

[0088] When the second recommendation effect is higher than the first recommendation effect, the result shows that conducting user behavior evaluation on the independent station store reduces the recommendation effect. Therefore, the historical recommendation effect result is that the recommendation effect without conducting user behavior evaluation on the independent station store is better.

[0089] In this embodiment, by conducting regular online measurements to evaluate user behavior of independent station stores to see whether the recommendation effectiveness has been improved, the corresponding historical recommendation effectiveness results can be determined, which can ensure the reliability and accuracy of the historical recommendation effectiveness results, help improve the accuracy of product recommendations, and make it possible to achieve the expected recommendation results.

[0090] In a further embodiment, before step S1300, determining that each recalled candidate product is associated with the user behavior score of the target user, the following steps are included:

[0091] Step S3300: obtaining training samples and supervisory labels in a prepared training set, wherein the training samples include commodity feature information of commodities and user feature information of users, and the supervisory labels represent whether the user performs a target user behavior on the commodity;

[0092] The product feature information includes all pre-processed texts describing the product and / or statistical features of the product. The text describing the product may be any one or more of the product category, product label, product title, product attributes, etc.; the statistical features may be the cumulative number of any one or more of the product's historical exposure, clicks, purchases, etc.

[0093] The user characteristic information includes the pre-processed user portrait, and / or the description text of the product when the user performs the target user behavior on the product, which can be any one or more of the product category, product label, product title, product attributes, etc.; the user portrait includes the user's basic information and / or user behavior data, and the user's basic information usually includes any one or more of gender, age, occupation, purchasing power, geographic location, product brand preference, product category preference, product price preference, etc.; the user behavior data includes statistics on specified interactive behaviors between users and products and / or independent station stores, such as the time period and / or frequency of users visiting independent station stores, and the frequency and / or time period of users browsing products, searching products, adding to shopping carts, purchasing products, clicking products, collecting products, and commenting on products in independent station stores.

[0094] The preprocessing is to format the processed information into a format suitable for model processing, which can be any one or more of one-hot encoding, feature normalization, regularization, etc., and those skilled in the art can flexibly implement it.

[0095] As for the training samples and their supervisory labels prepared in advance in the training set, specifically, by monitoring the interaction between users and commodities in the independent station store, when the user performs the target user behavior on the commodity, the user feature information of the user and the commodity feature information of the commodity are collected as training samples, and the supervisory label of the training sample is marked as representing that the user performs the target user behavior on the commodity, which can be any one or any multiple of the click-through rate (CTR) of 1, the conversion rate (CVR) of 1, the click-through conversion rate (CTCVR) of 1, etc.; when the user does not perform the target user behavior on the commodity, the user feature information of the user and the commodity feature information of the commodity are collected as training samples, and the supervisory label of the training sample is marked as representing that the user does not perform the target user behavior on the commodity, which can be any one or any multiple of the click-through rate (CTR) of 0, the conversion rate (CVR) of 0, the click-through conversion rate (CTCVR) of 0, etc.

[0096] Step S3310: Input the training sample into the behavior prediction model to extract corresponding deep semantic features and obtain a behavior feature vector;

[0097] It is recommended to use ESMM in the selection of the behavior prediction model. Other variant models such as LR, DeepFM model, etc. that can achieve the same functions of this application can also be replaced by it.

[0098] In one embodiment, a behavior prediction model based on ESSM is adopted, and each supervision label in the training set includes the click rate, click conversion rate, and conversion rate corresponding to the corresponding training sample. The training samples are input into the behavior prediction model, and the shared embedding layer in the behavior prediction model maps each high-dimensional sparse feature in the training samples to a corresponding low-dimensional dense feature. Furthermore, the first branch is that the pooling layer in the click prediction tower in the behavior prediction model performs bitwise element addition operation on each low-dimensional dense feature on the user side to obtain a user comprehensive feature, and at the same time, performs bitwise element addition operation on each low-dimensional dense feature on the product side to obtain a product comprehensive feature, and then horizontally splices the user comprehensive feature and the product comprehensive feature to obtain a deep semantic feature representing the possibility of clicking, and its corresponding vectorized representation, i.e., a behavior feature vector for predicting the click-through rate; the second branch is that the pooling layer in the conversion prediction tower in the behavior prediction model performs bitwise element addition operation on each low-dimensional dense feature on the user side to obtain a user comprehensive feature, and at the same time, performs bitwise element addition operation on each low-dimensional dense feature on the product side to obtain a product comprehensive feature, and then horizontally splices the user comprehensive feature and the product comprehensive feature to obtain a deep semantic feature representing the possibility of conversion, and its corresponding vectorized representation, i.e., a behavior feature vector for predicting the conversion rate.

[0099] Step S3320: classify and map the behavior feature vector to predict the corresponding user behavior score;

[0100] Following the first branch, the multi-layer perceptron (MLP) in the click prediction tower in the behavior prediction model performs a nonlinear transformation on the behavior feature vector used to predict the click-through rate, and then activates the sigmoid or softmax function to obtain the predicted click-through rate; following the second branch, the multi-layer perceptron (MLP) in the conversion prediction tower in the behavior prediction model performs a nonlinear transformation on the behavior feature vector used to predict the conversion rate, and then activates the sigmoid or softmax function to obtain the predicted conversion rate. In addition, the behavior prediction model multiplies the predicted click-through rate and the predicted conversion rate to obtain the predicted click-through conversion rate. Therefore, the predicted click-through rate, the predicted conversion rate, and the predicted click-through conversion rate are the user behavior scores corresponding to the training samples.

[0101] Step S3330: Calculate the loss value corresponding to the user behavior score based on the supervision label, and update the behavior prediction model based on the loss value until the model converges to determine the user behavior score of the user who triggered the recall and is associated with the recalled product.

[0102] Call the preset cross entropy loss function, calculate the cross entropy loss values ​​corresponding to the predicted click rate, predicted conversion rate, and predicted click-to-purchase rate according to the supervisory labels of the training samples, and when each cross entropy loss value reaches a preset threshold, it indicates that the behavior prediction model has been trained to a convergence state, thereby terminating the behavior prediction model training; when any cross entropy loss value does not reach the preset threshold, it indicates that the behavior prediction model has not converged, and then implement a gradient update on the model according to the cross entropy loss value, usually by back propagation to correct the weight parameters of each link of the model so that the model is closer to convergence, and then continue to call other training samples and their supervisory labels to implement iterative training on the behavior prediction model until the model is trained to a convergence state. The preset threshold can be pre-set by a person skilled in the art as needed according to the disclosure herein.

[0103] In this embodiment, the behavior prediction model is trained until it reaches a convergence state, so that it can learn the ability to reliably and accurately predict user behavior scores including any one or more of click-through rate, conversion rate, and click-to-conversion rate based on the input user feature information and product feature information.

[0104] In a further embodiment, step S1400, selecting a plurality of recommended products based on the recommendation score ranking and pushing them to the target user, comprises the following steps:

[0105] Step S1401, sort and select multiple candidate products with relatively high recommendation scores as recommended products, and construct a recommended product list;

[0106] The recalled multiple candidate products are sorted in descending order of recommendation scores, and then the top N candidate products are selected as recommended products, and these recommended products are collected to form a recommended product list. The N can be set as needed by those skilled in the art.

[0107] Step S1402: Initialize and construct effectiveness distribution of each recommended product in the recommended product list according to the user behavior score of each recommended product in the recommended product list;

[0108] Based on the algorithm principle of Thompson sampling, firstly, the user behavior score of each recommended product in the recommended product list is obtained. The user behavior scores of these recommended products are all scored under the same effectiveness evaluation dimension.

[0109] In one embodiment, when the effectiveness evaluation dimension is to estimate the click-through rate corresponding to whether the user clicks on the product after being exposed to the product, for each recommended product, the total number of historical exposures of the recommended product is obtained, and the user behavior score of the recommended product, i.e., the click-through rate, is multiplied by the total number of historical exposures to obtain a click prior value as α in the effectiveness distribution, and the click prior value is subtracted from the total number of historical exposures to obtain a no-click prior value as β in the effectiveness distribution, and Beta(α, β) is constructed using the α and β as the effectiveness distribution.

[0110] In another embodiment, when the effectiveness evaluation dimension is the click-through conversion rate corresponding to estimating whether the user will make a conversion behavior on the product after clicking on the product, for each recommended product, the total number of historical clicks of the recommended product is obtained, the user behavior score of the recommended product, i.e., the click-through conversion rate, is multiplied by the total number of historical clicks to obtain a click-through conversion prior value as α in the effectiveness distribution, and the total number of historical clicks minus the click-through conversion prior value is subtracted to obtain a click-through non-conversion prior value as β in the effectiveness distribution, and Beta (α, β) is constructed using α and β as the effectiveness distribution. The conversion behavior may be adding the product to a shopping cart and / or purchasing the product.

[0111] In another embodiment, when the effectiveness evaluation dimension is a conversion rate corresponding to predicting whether the user will make a conversion behavior on the product after being exposed to the product, for each recommended product, the total number of historical exposures of the recommended product is obtained, and the user behavior score of the recommended product, i.e., the conversion rate, is multiplied by the total number of historical exposures to obtain a conversion prior value as α in the effectiveness distribution, and the conversion prior value is subtracted from the total number of historical exposures to obtain a non-conversion prior value as β in the effectiveness distribution, and Beta(α, β) is constructed using the α and β as the effectiveness distribution.

[0112] In other embodiments, when the effectiveness evaluation dimension is a combination of any multiple of the following: the conversion rate corresponding to whether the user makes a conversion behavior on the product after the product is exposed, the click-through rate corresponding to whether the user clicks on the product after the product is exposed, and the click-through conversion rate corresponding to whether the user makes a conversion behavior on the product after clicking on the product, etc., for each recommended product, the total number of historical exposures of the recommended product is obtained, and the user behavior score of the recommended product, i.e., the value obtained by combining the corresponding items, is multiplied by the total number of historical exposures to obtain the effectiveness prior value as α in the effectiveness distribution, and the effectiveness prior value is subtracted from the total number of historical exposures to obtain the non-effectiveness prior value as β in the effectiveness distribution, and Beta (α, β) is constructed using the α and β as the effectiveness distribution. The value obtained by combining the multiple items, i.e., the user behavior score, can be obtained by weighted fusion of the corresponding items, i.e., each item is multiplied by its preset weight and then added together, and the weights of each item can be preset and set by technicians in this field as needed.

[0113] Step S1403: updating each achievement distribution according to the historical user behavior of the target user and each user in the user group to which the target user belongs and corresponding to the recommended product list;

[0114] In one embodiment, a clustering algorithm is used to cluster all users in the independent station store based on user feature information of all users, so as to determine at least one user group and the user group to which each user belongs. The clustering algorithm can be any one of the K-Means algorithm, DBSCAN algorithm, etc., and technical personnel in this field can flexibly implement it.

[0115] The historical user behavior corresponding to the target user and each user in the user group to which he belongs in the independent station store is associated with the recommended product list. It refers to whether the user has made a behavior corresponding to the effectiveness evaluation dimension for the recommended product in the recommended product list after the recommended product in the recommended product list has been historically exposed to the user, for the sake of understanding, an exemplary example is given. If the effectiveness evaluation dimension of the effectiveness distribution is to estimate the click-through rate corresponding to whether the user clicks on the product after the product is exposed, the corresponding historical user behavior is whether the user clicks on the recommended product. Thus, further, all the historical user behaviors of the target user and each user in the user group to which he belongs are obtained. For each historical user behavior, if the historical user behavior is a behavior corresponding to the effectiveness evaluation dimension of the effectiveness distribution, add one to α in the effectiveness distribution; if the historical user behavior is not a behavior corresponding to the effectiveness evaluation dimension of the effectiveness distribution, add one to β in the effectiveness distribution. In this way, real-time updating of the effectiveness distribution is achieved.

[0116] Step S1404: randomly sampling the corresponding expected effectiveness probabilities for the various effectiveness distributions, reordering the recommended products associated with the various effectiveness distributions according to the expected effectiveness probabilities corresponding to the various effectiveness distributions, and placing them on the recommendation display page of the target recommendation scenario;

[0117] For each achievement distribution, a random number is randomly generated as the expected probability of achievement based on its deformation parameters α and β using the achievement distribution.

[0118] The recommended products in the recommended product list are reordered in descending order of expected probability of effectiveness, and then a recommendation display page for a target recommendation scenario is obtained, and the ordered recommended products are placed in the page in order, that is, the higher the ordered recommended products are, the closer they are to the top of the page, so that users can give priority to browsing products with a higher expected probability of effectiveness based on their regular top-down browsing habits.

[0119] Step S1405: Push the recommendation display page to the target user to display the recommended products to the target user.

[0120] The recommendation display page is pushed to the target user, and the page is loaded and displayed, so that the recommended products are exposed to the target user, thereby achieving product recommendation for the target user.

[0121] In this embodiment, recommended products are selected based on the recommendation scores to construct a recommended product list. For each recommended product in the list, the effectiveness distribution of the recommended product is initialized based on the user behavior score of the target user as prior experience knowledge. On this basis, the effectiveness distribution of the recommended product is updated based on the historical user behavior of the target user and each user in the user group to which he belongs as a posteriori actual data. Finally, based on the expected probability of effectiveness obtained by random sampling of the latest effectiveness distribution of each recommended product, each recommended product is re-sorted and placed on the corresponding display page of the recommendation scenario and pushed to the target user, so that the recommendation results are more in line with the user's real-time preferences and behavior trends, which not only improves the accuracy of the recommendation, but also enhances the user experience, making it more likely that users will see the products they are interested in.

[0122] In a further embodiment, step S1400, selecting a plurality of recommended products based on the recommendation score ranking and pushing them to the target user, comprises the following steps:

[0123] Step S1410: sort and select multiple candidate products with relatively high recommendation scores as recommended products, and construct a recommended product list;

[0124] The recalled multiple candidate products are sorted in descending order of recommendation scores, and then the top N candidate products are selected as recommended products, and these recommended products are collected to form a recommended product list. The N can be set as needed by those skilled in the art.

[0125] Step S1420: determining a recommendation interaction heat map corresponding to the recommendation display page of the target recommendation scenario according to the historical interaction data of the target user and each user in the user group to which the target user belongs;

[0126] In one embodiment, a clustering algorithm is used to cluster all users in the independent station store based on user feature information of all users, so as to determine at least one user group and the user group to which each user belongs. The clustering algorithm can be any one of the K-Means algorithm, DBSCAN algorithm, etc., and technical personnel in this field can flexibly implement it.

[0127] Monitor the target user and each user in the user group to which he belongs, and the historical interaction data generated by each interaction with any recommended product in the recommendation display page of the target recommendation scene, and obtain these historical interaction data, wherein the historical interaction data includes the location information of the recommended product in the recommendation display page. The interaction includes any one or more of long press, click, add to cart, purchase, etc.

[0128] It can be understood that the specific location where the corresponding interaction occurred can be located in the recommendation display page according to each historical interaction data. For each location, the number of historical interaction data of the location is accumulated as the interaction frequency of the location, and then a recommendation interaction heat map of the same size as the recommendation display page is constructed. In this map, the color of each location is marked according to the interaction frequency of the location, so that the more the interaction frequency is accumulated, the darker the corresponding marked color is. In this way, the recommendation interaction heat map is obtained.

[0129] Step S1430: determining the interactive hot area in the recommendation display page according to the recommendation interaction heat map;

[0130] According to the color depth in the recommendation interaction heat map, an area formed by at least one position with a darker color than other positions is determined. The area is an area where interactions occur frequently, so the area is used as the high-heat interaction area in the recommendation display page.

[0131] Step S1440: Determine a purchasing power range according to the historical purchase data of the target user, and select recommended products that belong to the purchasing power range and have a high recommendation score from the recommended product list as regional display products;

[0132] The target user's historical purchase data is obtained, and the historical purchase data includes the payment amounts of all orders of the target user to purchase goods in the independent station store. All payment amounts are sorted in descending order according to the payment amount, and multiple payment amounts that are ranked first and last are filtered to ensure the reliability of purchasing power. Then, from all the filtered payment amounts, the corresponding payment amounts at the top and the bottom of the ranking are screened out, and the corresponding payment amounts are used as the maximum value and the minimum value in the value range, and the value range is determined as the purchasing power range.

[0133] From the recommended product list, the recommended products belonging to the purchasing power range are screened out, and then from these recommended products, recommended products with higher recommendation scores than other recommended products are screened out as regional display products.

[0134] Step S1450: place the products displayed in each area in the interactive high-heat area of ​​the recommended display page, place the remaining recommended products in the recommended product list in the non-interactive high-heat area of ​​the recommended display page, and push the recommended display page to the target user to display the recommended products to the target user.

[0135] The products displayed in each area are placed in the interactive hot area of ​​the recommendation display page according to the order of recommendation scores from high to low, that is, the recommended products with higher recommendation scores are closer to the top of the interactive hot area, so that users can browse products with higher recommendation scores preferentially based on their regular top-down browsing habits. In addition, the recommended products in the recommended product list other than those selected as regional display products are also placed in the non-interactive hot area outside the interactive hot area of ​​the recommendation display page according to the order of recommendation ranking scores from high to low. Thus, the recommendation display page with completed placement is pushed to the target user to recommend products to the target user.

[0136] In this embodiment, recommended products are selected based on the recommendation scores to construct a recommended product list, and based on the historical interaction data of the target user and each user in the user group to which he belongs, a recommendation interaction heat map corresponding to the recommendation display page of the target recommendation scenario and the interaction high-heat area in the map are determined. Accordingly, recommended products that match the purchasing power range of the target user are selected from the recommended product list and placed in the interaction high-heat area of ​​the recommendation display page, and the remaining unselected recommended products are placed in the non-interaction high-heat area of ​​the page. The resulting recommendation display page is pushed to the target user, making it easier for the target user to pay attention to the recommended products within his purchase range, thereby promoting the transaction of recommended products and improving the recommendation effectiveness.

[0137] See also Figure 3 , a product recommendation device provided to meet one of the purposes of the present application is a functional embodiment of the product recommendation method of the present application. On the other hand, the device is a product recommendation device provided to meet one of the purposes of the present application, including an event response module 1100, a product recall module 1200, a product evaluation module 1300 and a product push module 1400, wherein the event response module 1100 is used to respond to the product recommendation event of the independent station store and determine the target recommendation scenario and target user corresponding to the event; the product recall module 1200 is used to obtain the candidate product set respectively recalled by multiple recall sources corresponding to the target recommendation scenario and the initial recall of each candidate product A recall score is determined by adjusting the initial recall score of each candidate product in each candidate product set according to the effectiveness weight corresponding to each recall source compared with other recall sources to determine a target recall score; a product evaluation module 1300 is used to obtain the dynamically generated historical recommendation effectiveness results of the independent station store, and when the historical recommendation effectiveness results indicate that a user behavior evaluation of the store is required, the user behavior score of the target user associated with each recalled candidate product is determined; a product push module 1400 is used to determine a recommendation score according to the target recall score and user behavior score of each candidate product, and select a plurality of recommended products based on the recommendation score sorting and push them to the target user.

[0138] In a further embodiment, after the product evaluation module 1300, it includes: a first score determination sub-module, which is used to use the target recall score of each recalled candidate product as the recommendation score when the historical recommendation effectiveness results indicate that user behavior evaluation of the store is not required; and a first product push sub-module, which is used to screen out multiple recommended products whose recommendation scores meet preset conditions and push them to the target user.

[0139] In a further embodiment, before the event response module 1100, an effectiveness weight generation process is started, including: a recall source determination submodule, which is used to determine multiple recall sources associated with each recommendation scenario for each preset multiple recommendation scenarios corresponding to the independent station store; a second product push submodule, which is used to recall the corresponding target product for the user's product recommendation event based on each recall source under each recommendation scenario, and push the target products recalled by each recall source to the user; a weight determination submodule, which is used to obtain the effectiveness data generated by exposing the target product to the corresponding user, and for each recommendation scenario, determine the effectiveness weight corresponding to each recall source based on the effectiveness data of the target products recalled by each recall source under the recommendation scenario.

[0140] In a further embodiment, the product evaluation module 1300 includes: an event response submodule, which is used to respond to the timed triggered online recommendation effectiveness measurement event, and divide the online user recommendation flow associated with the same recommendation scenario into a first user recommendation flow and a second user recommendation flow; a first effectiveness determination submodule, which is used to recall the target products and their target recall scores of the corresponding recall sources for the product recommendation events of each user included in the first user recommendation flow, sort and select the recommended products based on the target recall score and push them to the user, then obtain the effectiveness data generated by the exposure of the recommended products to the corresponding user, and determine the first recommendation effectiveness based on all the effectiveness data; a second effectiveness determination submodule, which is used to determine the first recommendation effectiveness for the second user The user recommendation traffic includes product recommendation events for each user, and the target products and target recall scores of the corresponding recall sources are recalled, and the user behavior scores of the corresponding users associated with each target product are determined. After the recommended products are sorted and screened based on the target recall score and the user behavior score and pushed to the user, the effectiveness data generated by the exposure of the recommended products to the corresponding user is obtained, and the second recommendation effectiveness is determined based on all the effectiveness data; the result determination submodule is used to confirm that the historical recommendation effectiveness result is a better recommendation effectiveness for performing user behavior evaluation on the independent station store when the first recommendation effectiveness is higher than the second recommendation effectiveness, otherwise, confirm that the historical recommendation effectiveness result is a better recommendation effectiveness for not performing user behavior evaluation on the independent station store.

[0141] In a further embodiment, the product evaluation module 1300 includes: a data acquisition submodule for acquiring training samples and supervisory labels in a prepared training set, wherein the training samples include product feature information of the product and user feature information of the user, and the supervisory label represents whether the user performs a target user behavior on the product; a feature representation submodule for inputting the training samples into the behavior prediction model to extract corresponding deep semantic features and obtain a behavior feature vector; a classification mapping submodule for classifying and mapping the behavior feature vector to predict a corresponding user behavior score; a second score determination submodule for calculating a loss value corresponding to the user behavior score based on the supervisory label, and updating the behavior prediction model based on the loss value until the model converges, so as to determine the user behavior score of the user associated with the recalled product that triggers the recall.

[0142] In a further embodiment, the product push module 1400 includes: a list construction submodule, which is used to sort and screen out multiple candidate products with relatively high recommendation scores as recommended products, and construct a recommended product list; a distribution initialization submodule, which is used to initialize and construct the effectiveness distribution of each recommended product in the recommended product list according to the user behavior score of each recommended product in the recommended product list; a distribution update submodule, which is used to update each effectiveness distribution according to the historical user behavior corresponding to the target user and each user in the user group to which he belongs and associated with the recommended product list; a first page setting submodule, which is used to randomly sample the corresponding expected effectiveness probability for each effectiveness distribution, re-sort the recommended products associated with each effectiveness distribution according to the expected effectiveness probability corresponding to each effectiveness distribution, and place them in the recommendation display page of the target recommendation scenario; a first product recommendation submodule, which is used to push the recommendation display page to the target user to display the recommended products to the target user.

[0143] In a further embodiment, the product push module 1400 includes: a list construction submodule, which is used to sort and screen out multiple candidate products with relatively high recommendation scores as recommended products, and build a recommended product list; a heat map determination submodule, which is used to determine the recommendation interaction heat map corresponding to the recommendation display page of the target recommendation scenario according to the historical interaction data of the target user and each user in the user group to which he belongs; an area determination submodule, which is used to determine the interactive high-heat area in the recommendation display page according to the recommendation interaction heat map; a product screening submodule, which is used to determine the purchasing power range according to the historical purchase data of the target user, and screen out recommended products that belong to the purchasing power range and have high recommendation scores from the recommended product list as regional display products; a second product recommendation submodule, which is used to place each regional display product in the interactive high-heat area of ​​the recommendation display page, place the remaining recommended products in the recommended product list in the non-interactive high-heat area of ​​the recommendation display page, and push the recommendation display page to the target user to display the recommended products to the target user.

[0144] In order to solve the above technical problems, the present application also provides a computer device. Figure 4 As shown, a schematic diagram of the internal structure of a computer device. The computer device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. Among them, the computer-readable storage medium of the computer device stores an operating system, a database, and computer-readable instructions. The database may store a control information sequence. When the computer-readable instructions are executed by the processor, the processor may implement a product recommendation method. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. The memory of the computer device may store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor may execute the product recommendation method of the present application. The network interface of the computer device is used to connect and communicate with a terminal. Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0145] In this embodiment, the processor is used to execute Figure 3The memory stores the program code and various data required to execute the above modules or submodules. The network interface is used to transmit data between user terminals or servers. The memory in this embodiment stores the program code and data required to execute all modules / submodules in the product recommendation device of this application, and the server can call the program code and data of the server to execute the functions of all submodules.

[0146] The present application also provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the product recommendation method of any embodiment of the present application.

[0147] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments of the present application can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a computer-readable storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0148] To sum up, this application can serve various personalized recommendation scenarios of independent station stores, evaluate the recall value of products according to the recommendation scenarios, and timely evaluate the association value between products and users as needed, so as to recommend products that meet their expectations to users, attract users' attention, and hopefully facilitate the final product transaction and achieve high-quality recommendation results.

[0149] It will be understood by those skilled in the art that the various operations, methods, steps, measures, and schemes in the processes discussed in this application may be alternated, changed, combined, or deleted. Furthermore, other steps, measures, and schemes in the various operations, methods, and processes discussed in this application may also be alternated, changed, rearranged, decomposed, combined, or deleted. Furthermore, the steps, measures, and schemes in the various operations, methods, and processes in the prior art that are open source and in this application may also be alternated, changed, rearranged, decomposed, combined, or deleted.

[0150] The above description is only a partial implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A product recommendation method, characterized in that: The steps include: Respond to the product recommendation event of the independent station store and determine the target recommendation scenario and target user corresponding to the event; Obtaining candidate product sets respectively recalled by multiple recall sources corresponding to the target recommendation scenario and the initial recall score of each candidate product therein, and adjusting the initial recall score of each candidate product in each candidate product set according to the effectiveness weight corresponding to each recall source compared with other recall sources to determine the target recall score; Obtaining dynamically generated historical recommendation results of the independent website store, and when the historical recommendation results indicate that a user behavior evaluation of the store is required, determining a user behavior score of the target user associated with each recalled candidate product; A recommendation score is determined according to the target recall score and the user behavior score of each candidate product, and a plurality of recommended products are selected based on the recommendation score sorting and pushed to the target user.

2. The commodity recommendation method according to claim 1, characterized in that: After obtaining the dynamically generated historical recommendation results of the independent station store, the following steps are included: When the historical recommendation effectiveness results indicate that it is not necessary to conduct a user behavior evaluation on the store, the target recall score of each recalled candidate product is used as the recommendation score; A plurality of recommended products whose recommendation scores meet preset conditions are screened out and pushed to the target user.

3. The commodity recommendation method according to claim 1, characterized in that: Before responding to the product recommendation event of the independent station store, start the performance weight generation process, including the following steps: For the various preset recommendation scenarios corresponding to the independent station store, determine the multiple recall sources associated with each recommendation scenario; Based on each recall source in each recommendation scenario, recall corresponding target products for the user's product recommendation event, and push the target products recalled by each recall source to the user; The effectiveness data generated by exposing the target product to the corresponding user is obtained, and for each recommendation scenario, the effectiveness weight corresponding to each recall source is determined based on the effectiveness data of the target product recalled by each recall source in the recommendation scenario.

4. The commodity recommendation method according to claim 1, characterized in that: Obtaining the dynamically generated historical recommendation results of the independent website store includes the following steps: In response to the timed triggered online recommendation effectiveness measurement event, the online user recommendation traffic associated with the same recommendation scenario is divided into the first user recommendation traffic and the second user recommendation traffic; For the product recommendation events of each user included in the first user recommendation flow, the target products of each corresponding recall source and their target recall scores are recalled, and after the recommended products are sorted and screened based on the target recall scores and pushed to the user, the effectiveness data generated by the exposure of the recommended products to the corresponding user is obtained, and the first recommendation effectiveness is determined based on all the effectiveness data; For the product recommendation events of each user included in the second user recommendation flow, the target products of each corresponding recall source and their target recall scores are recalled, and the user behavior score of the corresponding user associated with each target product is determined. After the recommended products are sorted and screened based on the target recall score and the user behavior score and pushed to the user, the effectiveness data generated by the exposure of the recommended products to the corresponding user is obtained, and the second recommendation effectiveness is determined based on all the effectiveness data; When the first recommendation effect is higher than the second recommendation effect, it is confirmed that the historical recommendation effect result is that the recommendation effect of conducting user behavior evaluation on the independent station store is better; otherwise, it is confirmed that the historical recommendation effect result is that the recommendation effect of not conducting user behavior evaluation on the independent station store is better.

5. The commodity recommendation method according to claim 1, characterized in that: Before determining that each candidate product to be recalled is associated with the user behavior score of the target user, the following steps are included: Obtain training samples and supervisory labels in a prepared training set, wherein the training samples include commodity feature information of commodities and user feature information of users, and the supervisory labels represent whether the user performs a target user behavior on the commodity; Inputting the training sample into the behavior prediction model to extract corresponding deep semantic features and obtain a behavior feature vector; Classify and map the behavior feature vector to predict the corresponding user behavior score; A loss value corresponding to the user behavior score is calculated according to the supervisory label, and the behavior prediction model is updated according to the loss value until the model converges, so as to determine the user behavior score of the user who triggered the recall and is associated with the recalled product.

6. The commodity recommendation method according to claim 1, characterized in that: The method of selecting a plurality of recommended products based on the recommendation score ranking and pushing them to the target user comprises the following steps: Sorting and selecting multiple candidate products with relatively high recommendation scores as recommended products, and constructing a recommended product list; Initialize and construct effectiveness distribution of each recommended product in the recommended product list according to the user behavior score of each recommended product in the recommended product list; updating each achievement distribution according to the historical user behavior of the target user and each user in the user group to which the target user belongs in association with the recommended product list; Randomly sampling the corresponding expected effectiveness probabilities for each effectiveness distribution, reordering the recommended products associated with each effectiveness distribution according to the expected effectiveness probabilities corresponding to each effectiveness distribution, and placing them on the recommendation display page of the target recommendation scenario; The recommendation display page is pushed to the target user to display the recommended products to the target user.

7. The commodity recommendation method according to claim 1, characterized in that: The method of selecting a plurality of recommended products based on the recommendation score ranking and pushing them to the target user comprises the following steps: Sorting and selecting multiple candidate products with relatively high recommendation scores as recommended products, and constructing a recommended product list; Determine a recommendation interaction heat map corresponding to the recommendation display page of the target recommendation scenario according to the historical interaction data of the target user and each user in the user group to which the target user belongs; Determine the interactive hot area in the recommendation display page according to the recommendation interaction heat map; Determine a purchasing power range according to the historical purchase data of the target user, and select recommended products that belong to the purchasing power range and have a high recommendation score from the recommended product list as regional display products; The products displayed in each area are placed in the interactive high-heat area of ​​the recommended display page, the remaining recommended products in the recommended product list are placed in the non-interactive high-heat area of ​​the recommended display page, and the recommended display page is pushed to the target user to display the recommended products to the target user.

8. A commodity recommendation device, characterized in that: include: The event response module is used to respond to the product recommendation event of the independent station store and determine the target recommendation scenario and target user corresponding to the event; A product recall module is used to obtain the candidate product sets respectively recalled by multiple recall sources corresponding to the target recommendation scenario and the initial recall score of each candidate product therein, and adjust the initial recall score of each candidate product in each candidate product set according to the effectiveness weight corresponding to each recall source compared with other recall sources to determine the target recall score; A product evaluation module is used to obtain dynamically generated historical recommendation results of the independent website store, and when the historical recommendation results indicate that a user behavior evaluation of the store is required, determine the user behavior score of the target user associated with each recalled candidate product; The product push module is used to determine a recommendation score according to the target recall score and the user behavior score of each candidate product, and select multiple recommended products based on the recommendation score sorting and push them to the target user.

9. A computer device comprising a central processing unit and a memory, characterized in that: The central processing unit is used to call and run the computer program stored in the memory to execute the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: It stores a computer program implemented according to the method described in any one of claims 1 to 7 in the form of computer-readable instructions, and when the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

Images