Recommendation strategy dynamic adjustment method and its device, equipment and medium

Abstract

The present application relates to a method for dynamically adjusting recommendation strategies in the field of e-commerce technology and its device, equipment, and medium, the method comprising: determining a target recommendation scenario associated with an independent station store, obtaining a recommendation strategy set corresponding to the scenario based on a recommendation phase strategy library; starting an iterative process, and after applying each recommendation strategy in the recommendation strategy set online, determining the recommendation strategy with the highest measured recommendation effectiveness score as the preferred recommendation strategy; reconstructing a new recommendation strategy corresponding to the target recommendation scenario and higher than the latest highest recommendation effectiveness score based on the recommendation phase strategy library, replacing all other recommendation strategies except the preferred recommendation strategy in the recommendation strategy set, and continuing the iterative process; when a new recommendation strategy cannot be constructed, terminating the iterative process, and obtaining the latest preferred recommendation strategy for the target recommendation scenario as a practical recommendation strategy. The present application can fully explore the optimal recommendation strategy for a scenario.

Description

Technical Field

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

[0002] In the field of e-commerce, the importance of product recommendations is self-evident. It can significantly improve user experience and increase sales. However, the shortcoming of existing technologies is that they adopt a "one-size-fits-all" strategy and use the same recommendation strategy for all users or stores. This approach ignores the significant differences between different stores. Although they can provide a certain degree of recommendation services, they usually fail to fully consider the uniqueness of each independent station store. For example, different stores may have different target customer groups, product features, brand positioning and marketing strategies. This "one-size-fits-all" approach may lead to a lack of targeted recommendation results and fail to meet the specific needs of different stores, thus affecting the recommendation effect and user satisfaction.

[0003] 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

[0004] The primary purpose of the present application is to solve at least one of the above problems and to provide a method for dynamically adjusting a recommendation strategy and its corresponding device, computer equipment, and computer program product.

[0005] In order to meet the various objectives of this application, this application adopts the following technical solutions:

[0006] A method for dynamically adjusting a recommendation strategy provided to meet one of the purposes of this application includes the following steps:

[0007] Determine a target recommendation scenario from various recommendation scenarios associated with the independent station store, and obtain a recommendation strategy set corresponding to the target recommendation scenario based on the recommendation stage strategy library, wherein the recommendation strategy set includes multiple recommendation strategies;

[0008] Start the iterative process, apply each recommendation strategy in the recommendation strategy set online, and then determine the recommendation strategy with the highest measured recommendation effectiveness score as the preferred recommendation strategy;

[0009] Reconstructing a new recommendation strategy for the target recommendation scenario based on the recommendation phase strategy library, replacing all recommendation strategies other than the preferred recommendation strategy in the recommendation strategy set with the new recommendation strategy, and continuing the iterative process, wherein the predicted recommendation effectiveness score of the new recommendation strategy is higher than the measured recommendation effectiveness score of the preferred recommendation strategy;

[0010] When the new recommendation strategy cannot be constructed, the iterative process is terminated, and the latest preferred recommendation strategy for the target recommendation scenario is obtained as a practical recommendation strategy for product recommendations in the target recommendation scenario.

[0011] On the other hand, a recommendation strategy dynamic adjustment device provided to meet one of the purposes of the present application includes a set acquisition module, an iteration start module, an iteration continuation module and an iteration end module, wherein the set acquisition module is used to determine a target recommendation scenario from various recommendation scenarios associated with an independent station store, and obtain a recommendation strategy set constructed corresponding to the target recommendation scenario based on a recommendation stage strategy library, wherein the recommendation strategy set includes multiple recommendation strategies; the iteration start module is used to start the iteration process, and after applying each recommendation strategy in the recommendation strategy set online, the recommendation strategy with the highest measured recommendation effectiveness score is determined as the preferred recommendation strategy; the iteration continuation module is used to reconstruct a new recommendation strategy for the target recommendation scenario based on the recommendation stage strategy library, replace all recommendation strategies other than the preferred recommendation strategy in the recommendation strategy set with the new recommendation strategy, and continue the iteration process, wherein the predicted recommendation effectiveness score of the new recommendation strategy is higher than the measured recommendation effectiveness score of the preferred recommendation strategy; the iteration end module is used to end the iteration process when the new recommendation strategy cannot be constructed, and obtain the latest preferred recommendation strategy for the target recommendation scenario as a practical recommendation strategy for product recommendation in the target recommendation scenario.

[0012] 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 method for dynamically adjusting the recommendation strategy described in the present application.

[0013] 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.

[0014] The technical solution of this application has many advantages, including but not limited to the following aspects:

[0015] This application adapts the target recommendation scenario of the independent station store in the online environment, constructs a corresponding recommendation strategy set based on the recommendation stage strategy library and applies it, and through continuous iteration, determines the recommendation strategy with the highest measured recommendation effectiveness score of each recommendation strategy in the set as the preferred recommendation strategy, and dynamically replaces the non-preferred recommendation strategy, and introduces a new recommendation strategy with better prediction effect, so that after the iteration, it is ensured that the latest preferred recommendation strategy is the most practical at present, and serves as a practical recommendation strategy. It can be seen that by debugging the actual results of various recommendation strategies based on online dynamic debugging, various recommendation strategies can be continuously explored until the optimal recommendation strategy is finally selected, ensuring the real-time dynamic update of the recommendation strategy, and being actually reliable, so as to adapt to the characteristics of the independent station store and the characteristics of the market or user needs. Secondly, it has the adaptability of the recommendation scenario, ensuring the accuracy of the independent station store's recommendation of goods in the recommendation scenario, which helps to improve the recommendation effectiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] 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:

[0017] Figure 1 The network architecture of the e-commerce platform exemplified in this application;

[0018] Figure 2 A flowchart of a typical embodiment of the method for dynamically adjusting the recommendation strategy of the present application;

[0019] Figure 3 This is a principle block diagram of the recommended strategy dynamic adjustment device of this application;

[0020] Figure 4 A schematic diagram of the structure of a computer device used in this application. DETAILED DESCRIPTION

[0021] 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.

[0022] 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.

[0023] 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.

[0024] 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.

[0025] 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.

[0026] 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.

[0027] 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.

[0028] 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.

[0029] 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.

[0030] 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.

[0031] 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.

[0032] 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.

[0033] 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.

[0034] With the support of various components of the business management engine instance, the e-commerce platform 82 can provide online shopping functions, enabling 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 services, which are 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 services, which can define email notifications; API fulfillment services, which can trigger third-party applications to create fulfillment records in third parties; legacy fulfillment services, which can trigger custom API calls from the business management engine to third parties; gift card fulfillment services. You 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.

[0035] 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.

[0036] A method for dynamically adjusting a recommendation strategy 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.

[0037] See also Figure 2 The method for dynamically adjusting the recommendation strategy of the present application, in its typical embodiment, comprises the following steps:

[0038] Step S1100: determine a target recommendation scenario from various recommendation scenarios associated with independent website stores, and obtain a recommendation strategy set corresponding to the target recommendation scenario based on a recommendation stage strategy library, wherein the recommendation strategy set includes multiple recommendation strategies;

[0039] In order to determine a relatively small number of products to be recommended to users from the numerous products sold in independent station stores, the recommendation process can be divided into multiple stages, usually four stages in order of first arrival, namely recall, rough sorting, fine sorting, and re-sorting. It can also be other stages, for example, three stages in order of first arrival, namely recall, fine sorting, and filtering. The specific process can be determined by technical personnel in this field according to business needs.

[0040] The preset recommendation stage strategy library includes at least one stage strategy corresponding to each stage in the recommendation process and its applicable recommendation scenarios. The stage strategies corresponding to each stage can be set, for example: store high order recall, user region recall, similar product information recall, user behavior recall, etc. applicable to the recall stage; various fine ranking models with prediction tasks of CTR and / or CVR and / or CTCVR applicable to the fine ranking stage, such as: DeepFM, GDBT+LR, ListNet, etc.; novelty-based sorting, diversity-based sorting, display time-based sorting, etc. applicable to the re-ranking stage. Secondly, all the stage strategies in the recommendation stage strategy library can set their applicable recommendation scenarios. For example, for the stage strategies set in the recall stage, for the recommendation scenario of the user's click behavior, since the user clicks on the product because of interest, in order to achieve the recommendation that meets the user's expectations, the various stage strategies applicable to the recommendation scenario can 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 the recommendation scenario of the homepage, in order to show the user the products that may be of interest to the user, the various stage strategies applicable to the recommendation scenario can be user behavior recall, collaborative filtering recall, store high click recall, store high purchase recall, store high order recall, user region recall, etc., to recall the products of interest to the user; for the refined ranking stage and the filtering stage, the stage strategies of these stages are usually universal and applicable to various recommendation scenarios. Those skilled in the art can set the recommendation stage strategy library according to the disclosure here and the business needs.

[0041] In an independent station store, recommendation scenarios are pre-configured for various different product recommendation events that occur therein. The recommendation scenario may refer to the recommendation of the page and / or behavior that triggers the corresponding product recommendation event, or may refer to the recommendation of an activity formulated based on the operation and / or marketing situation of the independent station 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. In other words, the various recommendation scenarios associated with the independent station store are all recommendation scenarios that the independent station store can serve, and any one of these recommendation scenarios can be determined as the target recommendation scenario.

[0042] It is not difficult to understand that the recommendation stage library is used to construct a recommendation strategy for recommending products in a recommendation scenario. Specifically, all stage strategies applicable to each stage in the recommendation process and applicable to the target recommendation scenario in the recommendation stage library are determined as candidate stage strategies, each stage in the recommendation process is selected, and each candidate stage strategy of each stage is combined according to the sequence of these stages to obtain multiple candidate recommendation strategies, that is, a single candidate recommendation strategy includes a candidate stage strategy for each stage, and / or, some stages in the recommendation process are selected, for example: the various stages in the recommendation process are recall, rough sorting, fine sorting, and re-sorting, and recall and re-sorting are selected. Technical personnel in this field can set the selection of some stages of the recommendation process according to business needs, and then, each candidate stage strategy of each stage in the partial stages is combined according to the sequence of these stages to obtain multiple candidate recommendation strategies, that is, one candidate recommendation strategy includes a candidate stage strategy for each stage. Furthermore, a preset effectiveness prediction model is used to predict the recommendation effectiveness scores corresponding to the target recommendation scenarios of each candidate recommendation strategy, and all candidate recommendation strategies are sorted in descending order according to the predicted recommendation effectiveness scores, and multiple candidate recommendation strategies with the highest rankings are selected and added to the recommendation strategy set as recommendation strategies, and the recommendation strategy set is pre-initialized to be empty. The number of the selected candidate recommendation strategies can be set by those skilled in the art as needed.

[0043] The effectiveness prediction model is pre-trained to a convergence state and learns the ability to predict the recommendation effectiveness score corresponding to the product recommendation in the recommendation scenario to which the recommendation strategy is applicable. The specific training process is further disclosed in the subsequent embodiments, and this step is not discussed for the time being.

[0044] Step S1200, starting an iterative process, after applying each recommendation strategy in the recommendation strategy set online, determining the recommendation strategy with the highest measured recommendation effectiveness score as the preferred recommendation strategy;

[0045] Each recommendation strategy in the recommendation strategy set is deployed online and applied to the service of product recommendation. In one embodiment, whenever a user in an independent station store triggers a product recommendation event associated with the target recommendation scenario, the terminal device of the user carrying the independent station store constructs an online recommendation request carrying the event identifier of the product recommendation event and sends it to the server. The server receives each online recommendation request and responds to the event, and determines the recommendation scenario pre-configured for the event according to the unique identifier, that is, the target recommendation scenario. Further, these online recommendation requests are assigned to each recommendation strategy in the recommendation strategy set one by one in turn, so that each recommendation strategy is called to respond to the same number of online recommendation requests on average. Taking a single online recommendation request as an example, the corresponding single recommendation strategy is called to be applicable to the corresponding stage strategy of each stage in the recommendation process, and a product recommendation list consisting of a relatively small number of recommended products is determined from the numerous products in the independent station store and pushed to the user who triggered the online recommendation request to respond to the online recommendation request. After calling each recommendation strategy to answer each online recommendation request assigned to it, for each online recommendation request, by monitoring whether the corresponding user performs the target user behavior on the recommended products in the product recommendation list, determine the total number of recommended products corresponding to the target user behavior of the user among all the recommended products in the product recommendation list, and 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. When the total number of online recommendation requests answered by each recommendation strategy reaches the preset threshold, for each recommendation strategy, the total number of recommended products corresponding to the target user behavior of the user determined by the corresponding monitoring after each online recommendation request of the recommendation strategy is accumulated to obtain the total number of effective products, and the total number of recommended products corresponding to the target user behavior of the user is obtained to obtain the total number of ineffective products, and the ratio of the total number of effective products to the total number of ineffective products is used as the measured recommendation effectiveness score of the recommendation strategy. The target user behavior can be any one or more of click, add to shopping cart, purchase, and repeat purchase. The preset threshold can be flexibly set by those skilled in the art. Thus, the measured recommendation effectiveness score of each recommendation strategy in the recommendation strategy set can be obtained. All the recommendation strategies in the recommendation strategy set are sorted in descending order according to the measured recommendation effectiveness score, and the recommendation strategy with the highest ranking is selected as the preferred recommendation strategy.

[0046] Step S1300: reconstructing a new recommendation strategy for the target recommendation scenario based on the recommendation phase strategy library, replacing all recommendation strategies other than the preferred recommendation strategy in the recommendation strategy set with the new recommendation strategy, and continuing the iterative process, wherein the predicted recommendation effectiveness score of the new recommendation strategy is higher than the measured recommendation effectiveness score of the preferred recommendation strategy;

[0047] It can be understood that based on the recommendation stage strategy library, multiple candidate recommendation strategies suitable for the target recommendation scenario other than the recommendation strategies previously added to the recommendation strategy set can be reconstructed, and then the preset effectiveness prediction model is used to predict the recommendation effectiveness scores corresponding to the target recommendation scenarios for each of the candidate recommendation strategies, and all candidate recommendation strategies are sorted in order from high to low according to the predicted recommendation effectiveness scores, and multiple candidate recommendation strategies corresponding to the top rankings and predicted recommendation effectiveness scores higher than the actual recommended effectiveness scores of the preferred recommendation strategies are screened out as new recommendation strategies, and then all other recommendation strategies except the preferred recommendation strategy in the recommendation strategy set are replaced by these new recommendation strategies, and the iterative process continues. The number of reconstructions and screenings of the candidate recommendation strategies can be set as needed by those skilled in the art.

[0048] Step S1400: When the new recommendation strategy cannot be constructed, the iterative process ends, and the latest preferred recommendation strategy for the target recommendation scenario is obtained as a practical recommendation strategy for product recommendations in the target recommendation scenario.

[0049] It can be understood that if all candidate recommendation strategies other than the recommendation strategies previously added to the recommendation strategy set are reconstructed based on the recommendation stage strategy library, and then the preset effectiveness prediction model is used to predict the recommendation effectiveness scores of each of the candidate recommendation strategies, and all predicted recommendation effectiveness scores are lower than the actual recommendation effectiveness scores of the latest preferred recommendation strategy in the recommendation strategy set, then a new recommendation strategy cannot be constructed at this time. At this time, the preferred recommendation strategy in the recommendation strategy set is the latest preferred recommendation strategy, and there is no recommendation strategy better than the preferred recommendation strategy. Therefore, the iterative process is terminated, and the preferred recommendation strategy is used as a practical recommendation strategy to respond to online recommendation requests and provide corresponding product recommendation lists to the users who triggered the online recommendation requests, so as to provide product recommendation services to users of independent station stores.

[0050] 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:

[0051] This application adapts the target recommendation scenario of the independent station store in the online environment, constructs a corresponding recommendation strategy set based on the recommendation stage strategy library and applies it, and through continuous iteration, determines the recommendation strategy with the highest measured recommendation effectiveness score of each recommendation strategy in the set as the preferred recommendation strategy, and dynamically replaces the non-preferred recommendation strategy, and introduces a new recommendation strategy with better prediction effect, so that after the iteration, it is ensured that the latest preferred recommendation strategy is the most practical at present, and serves as a practical recommendation strategy. It can be seen that by debugging the actual results of various recommendation strategies based on online dynamic debugging, various recommendation strategies can be continuously explored until the optimal recommendation strategy is finally selected, ensuring the real-time dynamic update of the recommendation strategy, and being actually reliable, so as to adapt to the characteristics of the independent station store and the characteristics of the market or user needs. Secondly, it has the adaptability of the recommendation scenario, ensuring the accuracy of the independent station store's recommendation of goods in the recommendation scenario, which helps to improve the recommendation effectiveness.

[0052] In a further embodiment, step S1200, after applying each recommendation strategy in the recommendation strategy set online, determining the recommendation strategy with the highest measured recommendation effectiveness score as the preferred recommendation strategy includes the following steps:

[0053] Step S1210, calling each recommendation strategy in the recommendation strategy set to respectively respond to at least one online recommendation request associated with the target recommendation scenario, and determining a measured recommendation effectiveness score of each recommendation strategy;

[0054] In one embodiment, the server assigns all received online recommendation requests associated with the target recommendation scenario to each recommendation strategy in the recommendation strategy set in turn, so that each recommendation strategy is called to respond to the same number of online recommendation requests on average. Taking a single online recommendation request as an example, the corresponding single recommendation strategy is called to apply to the stage strategy corresponding to each stage in the recommendation process, and a product recommendation list consisting of a relatively small number of recommended products is determined from the numerous products in the independent station store and pushed to the user who triggered the online recommendation request to respond to the online recommendation request. After calling each recommendation strategy to respond to each online recommendation request assigned to it, for each online recommendation request, by monitoring whether the corresponding user performs the target user behavior on the recommended products in the product recommendation list, the total number of recommended products corresponding to the target user behavior performed by the user, and the total number of recommended products corresponding to the target user behavior not performed by the user, among all the recommended products in the product recommendation list, is determined. When the total number of online recommendation requests answered by each recommendation strategy reaches the preset threshold, for each recommendation strategy, the total number of recommended products corresponding to the target user behavior determined by the monitoring after each online recommendation request answered by the recommendation strategy is accumulated to obtain the total number of effective products, and the total number of recommended products corresponding to the target user behavior determined by the user is accumulated to obtain the total number of ineffective products, and the ratio of the total number of effective products to the total number of ineffective products is used as the measured recommendation effectiveness score of the recommendation strategy. The preset threshold can be set by those skilled in the art as needed.

[0055] In other alternative embodiments, each recommendation strategy can be first concentrated for the recommendation strategy. If the recommendation strategy is the preferred recommendation strategy, the measured recommendation effectiveness score of the recommendation strategy is obtained. If the recommendation strategy is not the preferred recommendation strategy, the predicted recommendation effectiveness score of the recommendation strategy is obtained. All recommendation strategies are sorted in order from low to high according to the recommendation effectiveness score. Then, the server allocates all the received online recommendation requests associated with the target recommendation scenario according to the ranking of each recommendation strategy, so that a higher-ranked recommendation strategy is called to respond to a larger number of online recommendation requests. The specific number of allocations and the implementation method can be flexibly implemented by those skilled in the art. In this way, recommendation strategies with lower recommendation effectiveness scores can have more opportunities to be called, which helps to reduce prediction errors.

[0056] Step S1220: calling the two recommendation strategies with higher measured recommendation scores to respectively respond to at least one online recommendation request associated with the recommendation scenario, determining that the two recommendation strategies correspond to new measured recommendation effectiveness scores, and updating the original measured recommendation scores;

[0057] Furthermore, all recommendation strategies in the recommendation strategy set are sorted in descending order according to the measured recommendation effectiveness scores, and the top two recommendation strategies are selected. Then, the server redirects and distributes all online recommendation requests related to the target recommendation scenario received, changing the original corresponding distribution to each recommendation strategy in the recommendation strategy set to the distribution to the two selected recommendation strategies one by one in turn, so that the two recommendation strategies respond to the same number of online recommendation requests on average. Afterwards, in the same manner as above, the new measured recommendation effectiveness scores corresponding to the two recommendation strategies are determined, and then the original measured recommendation scores are updated respectively.

[0058] Step S1230: Determine the preferred recommendation strategy as the recommendation strategy with the highest measured recommendation effectiveness score among the two recommendation strategies.

[0059] According to the latest measured recommendation effectiveness scores of the two recommendation strategies, the recommendation strategy with the highest recommendation effectiveness score is selected as the preferred recommendation strategy.

[0060] In this embodiment, first, by allowing each recommendation strategy in the recommendation strategy set to respond to the same number of online recommendation requests on average, it is ensured that each recommendation strategy has a fair opportunity to demonstrate its effectiveness, which helps to collect balanced measured data, thereby accurately evaluating the effectiveness of each recommendation strategy. Afterwards, by monitoring the user's response to the product recommendation list, the number of valid recommendations and invalid recommendations for each recommendation strategy can be accurately counted, and then the measured recommendation effectiveness score can be calculated. The scoring mechanism based on user behavior feedback makes the recommendation optimization closer to the actual needs of users. Finally, based on the two-stage optimization, that is, first preliminarily screening out two recommendation strategies with better results, and then finely screening out the recommendation strategy with the best results between the two, the overall optimization efficiency is higher while ensuring the accuracy and reliability of the optimization.

[0061] In a further embodiment, step S1300, reconstructing a new recommendation strategy for the target recommendation scenario based on the recommendation stage strategy library, comprises the following steps:

[0062] Step S1310: reconstruct at least one candidate recommendation strategy applicable to the recommendation scenario based on at least one corresponding stage strategy among the recommendation strategies set for each stage in the recommendation process in the recommendation stage strategy library;

[0063] Based on the recommendation phase strategy library, a plurality of candidate recommendation strategies applicable to the target recommendation scenario other than the recommendation strategies previously added to the recommendation strategy set can be reconstructed.

[0064] Step S1320: using a preset effectiveness estimation model to predict the recommendation effectiveness score of each candidate recommendation strategy;

[0065] The effectiveness prediction model is pre-trained to a convergence state and learns the ability to predict the recommendation effectiveness score corresponding to the product recommendation in the recommendation scenario to which the recommendation strategy is applicable. The specific training process is further disclosed in the subsequent embodiments, and this step is not discussed for the time being.

[0066] Step S1330: When the predicted recommendation effectiveness score of the candidate recommendation strategy is higher than the actual measured recommendation effectiveness score of the latest preferred recommendation strategy, the candidate recommendation strategy is confirmed as a new recommendation strategy.

[0067] All candidate recommendation strategies are sorted in descending order according to the predicted recommendation effectiveness scores, and multiple candidate recommendation strategies with high rankings and predicted recommendation effectiveness scores higher than the actual recommended effectiveness scores of the preferred recommendation strategy are selected as new recommendation strategies.

[0068] In this embodiment, by first reconstructing multiple candidate recommendation strategies, and using the effectiveness prediction model to predict the recommendation effectiveness score corresponding to each candidate recommendation strategy, and then selecting the candidate recommendation strategy whose predicted recommendation effectiveness score is higher than the latest measured recommendation effectiveness score as the new recommendation strategy, it can ensure that possible high-quality recommendation strategies are accurately excavated.

[0069] In a further embodiment, before step S1320, using a preset effectiveness prediction model to predict the recommendation effectiveness score of each candidate recommendation strategy, includes the following steps:

[0070] Step S2300: obtaining training samples and supervisory labels in a prepared training set, wherein the training samples include a historical recommendation strategy and its applicable recommendation scenario, and the supervisory label is a measured recommendation effectiveness score corresponding to the commodity recommendation of the historical recommendation strategy in the recommendation scenario;

[0071] It can be understood that the historical recommendation strategies that have been deployed and applied online in each independent station store implemented by the e-commerce platform can be collected in advance, and the recommendation scenarios served by each historical recommendation strategy when applied can be determined. Usually, the recommendation scenarios are applicable to the corresponding historical recommendation strategies. In addition, after each historical recommendation strategy is applied online to serve its applicable recommendation scenario and provide the user with a corresponding product recommendation list, the real-time feedback corresponding to each historical recommendation strategy is determined based on monitoring, that is, for each historical recommendation strategy, after accumulating the online recommendation request triggered by the historical recommendation strategy each time in response to its applicable recommendation scenario, the total number of recommended products corresponding to the target user behavior determined by the corresponding monitoring is used to obtain the total number of effective products, and the total number of recommended products corresponding to the target user behavior determined by the user is used to obtain the total number of ineffective products, and the ratio of the total number of effective products to the total number of ineffective products is used as the measured recommendation effectiveness score of the historical recommendation strategy.

[0072] Then, each historical recommendation strategy is associated with its applicable recommendation scenario to form a single training sample. The supervisory label of each training sample is annotated with the actual recommendation effectiveness score corresponding to the product recommendation of the applicable recommendation scenario used by the historical recommendation strategy of the training sample. Then, all training samples and their supervisory labels are collected to form a training set for future use.

[0073] Step S2310: Input the training sample into the effectiveness prediction model to extract corresponding deep semantic features to obtain an effectiveness feature vector;

[0074] The effectiveness prediction model can be a logistic regression model suitable for binary classification tasks, such as an FM model, a GDBT+LR model, an MLP model, an MLR model, etc., and those skilled in the art can flexibly select and implement it.

[0075] The hidden layer in the effectiveness prediction model extracts the features of the training samples that are related to the recommendation effectiveness, and obtains deep semantic features. In this process, the features can be obtained through low-order and / or high-order feature interactions, and / or sequence position feature interactions, and / or abstract feature representations, and / or multimodal feature fusion. Then, the output is a vectorized representation of the effectiveness feature vector corresponding to the deep semantic feature.

[0076] Step S2320: classify and map the effectiveness feature vector to predict the corresponding prediction recommendation effectiveness score;

[0077] The effectiveness feature vector is classified and mapped, usually to a probability range between zero and one, where the closer it is to zero, the less likely it is to achieve the recommended effect, and the closer it is to one, the more likely it is to achieve the recommended effect. Thus, the mapped probability value is obtained as the predicted recommendation effect score.

[0078] Step S2330, calculate the loss value corresponding to the predicted recommendation effectiveness score based on the supervision label, and update the effectiveness estimation model based on the loss value until the model converges, so as to predict the recommendation effectiveness score corresponding to the product recommendation of the recommendation strategy for its applicable recommendation scenario.

[0079] Call the preset cross entropy loss function, calculate the cross entropy loss value corresponding to the predicted recommendation effectiveness score according to the supervisory label of the training sample, when the cross entropy loss value reaches the preset threshold, it indicates that the effectiveness prediction model has been trained to a convergence state, so that the effectiveness prediction model training can be terminated; when the cross entropy loss value does not reach the preset threshold, it indicates that the effectiveness prediction model has not converged, so the model is gradient updated according to the cross entropy loss value, usually by back propagation to correct the weight parameters of each link of the model to make the model closer to convergence, and then continue to call other training samples and their supervisory labels to perform iterative training on the effectiveness 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.

[0080] In this embodiment, the effectiveness prediction model is trained until it reaches a convergence state, so that it learns the ability to reliably and accurately predict the recommendation effectiveness scores corresponding to the product recommendations of the recommendation strategies used in the applicable recommendation scenarios. It has applicability in multiple recommendation scenarios, that is, it can predict the corresponding recommendation effectiveness scores for the recommendation strategies called in various target recommendation scenarios.

[0081] In a further embodiment, after step S1400, obtaining the latest preferred recommendation strategy of the target recommendation scenario as the practical recommendation strategy, the following steps are included:

[0082] Step S1510, determining the total number of events corresponding to the decrease in the difference in the recommendation effectiveness score measured by the practical recommendation strategy exceeding a first preset threshold;

[0083] It can be understood that the practical recommendation strategy can be continuously monitored to determine its corresponding measured recommendation effectiveness score, and the specific implementation can be flexibly implemented by those skilled in the art.

[0084] The first preset threshold is used to measure whether the difference in the decrease in the recommendation effectiveness score is large enough. If the difference in the decrease in the recommendation effectiveness score actually measured by the practical recommendation strategy exceeds the first preset threshold, it means that the decrease in the recommendation effectiveness score is too large and exceeds expectations; if the difference in the decrease in the recommendation effectiveness score actually measured by the practical recommendation strategy is less than or equal to the first preset threshold, it means that the decrease in the recommendation effectiveness score is within the expected range and is not large enough; those skilled in the art can set the first preset threshold as needed according to the explanation here.

[0085] Step S1520: When the total number of times exceeds a second preset threshold, restart the iterative process to redetermine a new practical recommendation strategy.

[0086] The second preset threshold can be pre-set to 1, or a value greater than 1. If the second preset threshold is 1, it means that once it is confirmed that the actual measured recommendation effect of the practical recommendation strategy has decreased significantly, so that it is greater than the second preset threshold, then the iterative process will be restarted immediately, and a new practical recommendation strategy can be re-determined, so that it can be updated in time to the practical recommendation strategy with the best actual measured effect; if the second preset threshold is a value greater than 1, it means that the difference in the actual measured recommendation effect of the practical recommendation strategy is greater than the total number of times corresponding to the first preset threshold, and the range in which it is less than or equal to the second preset threshold is used as the appropriate fluctuation range for a significant decrease. When the total number exceeds the appropriate fluctuation range, the restart of the iterative process is triggered, and a new practical recommendation strategy can be re-determined, so that it can be updated in time to the practical recommendation strategy with the best actual measured effect. Those skilled in the art can set the second preset threshold as needed based on the disclosure herein.

[0087] In this embodiment, the process of dynamically adjusting the practical recommendation strategy is disclosed, which can respond to the changes in effectiveness in a timely manner, thereby providing more accurate recommendations and maintaining the most effective recommendations in order to maintain the user experience. In addition, by flexibly setting the preset threshold, it is also possible to balance stability and responsiveness, and adjust appropriately on demand to avoid wasting resources.

[0088] In a further embodiment, step S1200, after applying each recommendation strategy in the recommendation strategy set online, determining the recommendation strategy with the highest measured recommendation effectiveness score as the preferred recommendation strategy includes the following steps:

[0089] Step S1201: for each recommendation strategy in the recommendation strategy set, if the recommendation strategy is a preferred recommendation strategy, obtain the measured recommendation effectiveness score of the recommendation strategy; if the recommendation strategy is not a preferred recommendation strategy, obtain the predicted recommendation effectiveness score of the recommendation strategy; and construct the effectiveness distribution of each recommendation strategy based on the recommendation effectiveness score of each recommendation strategy, wherein the predicted recommendation effectiveness score is obtained for the corresponding recommendation strategy by using a preset effectiveness prediction model;

[0090] It is not difficult to understand that for the recommended strategy that is the preferred recommended strategy in the recommended strategy set, in the iterative process before this step, the measured recommendation effectiveness score of the recommended strategy has been determined, thereby obtaining the measured recommendation effectiveness score of the recommended strategy; for the recommended strategy that is not the preferred recommended strategy in the recommended strategy set, in the construction process before this step, the preset effectiveness prediction model has been used to predict the recommended strategy to obtain the predicted recommendation effectiveness score of the recommended strategy, thereby obtaining the predicted recommendation effectiveness score of the recommended strategy.

[0091] Then, based on the Thompson sampling algorithm principle, the recommendation effectiveness scores of each recommendation strategy in the recommendation strategy set are multiplied by the same preset call number, and the corresponding α in the effectiveness distribution of each recommendation strategy is obtained. The preset call number is subtracted from each α, and the corresponding β in the effectiveness distribution of each recommendation strategy is obtained. For the effectiveness distribution of each recommendation strategy, Beta (α, β) constructed by α and β in the effectiveness distribution is used as the effectiveness distribution. The preset call number can be flexibly set by those skilled in the art according to business needs.

[0092] Step S1202, start the sampling iteration process, randomly sample each of the effectiveness distributions to determine the corresponding expected effectiveness probability, call the recommendation strategy with the highest expected effectiveness probability and apply it online to multiple online recommendation requests associated with the target recommendation scenario, and then determine the total number of valid recommendations and the total number of invalid recommendations measured by the recommendation strategy, update the effectiveness distribution of the recommendation strategy according to the total number of valid recommendations and the total number of invalid recommendations, and continue the sampling iteration process;

[0093] For each effectiveness distribution, the effectiveness distribution is used, based on its deformation parameters α and β, to randomly generate a random number as the effectiveness expected probability. Then, from the effectiveness expected probabilities of each effectiveness distribution, the recommendation strategy with the highest effectiveness expected probability is determined, and the server assigns all multiple online recommendation requests to the recommendation strategy, so that the recommendation strategy is called to respond to these online recommendation requests, and the server will not call any other recommendation strategies before completing the response to these online recommendation requests.

[0094] For each online recommendation request, the stage strategy applicable to each stage of the recommendation process in the recommendation strategy is called, and a relatively small number of recommended products are determined from the numerous products of the independent station store to form a product recommendation list, which is pushed to the user who triggered the online recommendation request to respond to the online recommendation request. After calling the recommendation strategy to respond to each online recommendation request assigned to it, for each online recommendation request, by monitoring whether the corresponding user performs the target user behavior on the recommended products in the product recommendation list, the total number of recommended products corresponding to the user's target user behavior and the total number of recommended products corresponding to the user's failure to perform the target user behavior are determined among all the recommended products in the product recommendation list. When the recommendation strategy is called to complete the response to these online recommendation requests, the total number of recommended products corresponding to the user's target user behavior determined by the corresponding monitoring after each response to the online recommendation request by the recommendation strategy is accumulated to obtain the total number of effective products, which is the total number of effective recommendations, and the total number of recommended products corresponding to the user's failure to perform the target user behavior is obtained to obtain the total number of ineffective products, which is the total number of invalid recommendations. Then, α in the effectiveness distribution of the recommendation strategy is updated to the value of the sum of it and the total number of valid recommendations, and β in the effectiveness distribution is updated to the value of the sum of it and the total number of invalid recommendations, thereby completing the update of the effectiveness distribution and continuing the sampling iterative process.

[0095] Step S1203: When the number of sampling iterations exceeds a preset threshold, the sampling iteration process is terminated, and the measured recommendation effectiveness scores of each recommendation strategy are determined according to the latest effectiveness distributions of each recommendation strategy, and the recommendation strategy with the highest measured recommendation effectiveness score is determined as the preferred recommendation strategy.

[0096] Construct a counter with the number of sampling iterations initialized to zero, and add one to the number of sampling iterations after each execution of the sampling iteration process. When the counter is read to determine that the number of sampling iterations exceeds the preset threshold, the sampling iteration process ends, and for each of the latest effectiveness distributions of each recommendation strategy, the ratio between α and β in the effectiveness distribution is used as the measured recommendation effectiveness score of the recommendation strategy associated with the effectiveness distribution. Then, all the recommended strategies in the recommendation strategy set are sorted in descending order according to the measured recommendation effectiveness scores, and the recommended strategy with the highest ranking is selected as the preferred recommended strategy. The preset threshold can be set by a person skilled in the art as needed.

[0097] In this embodiment, by using the principle of Thompson sampling algorithm based on the recommendation effectiveness scores of each recommendation strategy in the recommendation strategy set, the effectiveness distribution of each recommendation strategy is initialized and constructed, and then by randomly sampling the effectiveness distribution of each recommendation strategy, the recommendation strategy with the highest expected probability of effectiveness is determined and applied to the actual online recommendation request. In this process, by monitoring the user's response to the product recommendation list, the total number of valid and invalid recommendations is accumulated, and the recommendation strategy is dynamically adjusted according to the effectiveness distribution of the real-time feedback update strategy, so that it can continuously learn and adapt to changes in user behavior. In the sampling iteration process, both exploration and utilization are taken into account to more accurately capture user preferences, thereby improving the relevance of recommendations and user satisfaction.

[0098] When the number of sampling iterations exceeds the preset threshold, the iteration process ends, and the optimal recommendation strategy is determined based on the measured recommendation strategies determined by the latest effectiveness distribution, ensuring that the optimal recommendation strategy is based on sufficient data and iterative optimization results, thereby improving the stability and reliability of the recommendation.

[0099] In a further embodiment, after step S1400, obtaining the latest preferred recommendation strategy of the target recommendation scenario as the practical recommendation strategy, the following steps are included:

[0100] Step S1501, responding to a recommendation strategy setting event of a new independent station store, determining that the new independent station store associated with the event belongs to a plurality of independent station stores corresponding to stores of the same type;

[0101] The newly-born independent station store may be a newly-created independent station store, or an independent station store whose operating time is less than a preset time. The preset time may be pre-set by technicians in this field according to the disclosure herein and business requirements.

[0102] The e-commerce platform provides a recommendation strategy setting service to each independent station store it implements. When any independent station store chooses to use this service, it can trigger the recommendation strategy setting event of the independent station store, thereby setting its practical recommendation strategy for the independent station store.

[0103] In one embodiment, for each independent station store implemented on the e-commerce platform, the store business characteristics such as the store business category, the main price of the store, and the store business label of the independent station store are collected. The store business category includes different categories to which the goods in the corresponding independent station store belong; the store business price range includes the price of the main goods in the corresponding independent station store; the store business label includes different labels to which the goods in the corresponding independent station store belong, and the label can be obtained by extracting keywords from the description text of the corresponding goods. The description text includes all texts describing the corresponding goods, such as the product title, product details text, product introduction text, etc., and those skilled in the art can flexibly and flexibly implement the determination of the business characteristics of each store. Further, a clustering algorithm or a deep learning algorithm is used to determine whether each two independent station stores belong to the same type of store based on the store business characteristics of each independent station store. The clustering algorithm can be any one of the K-Means algorithm, the DBSCAN algorithm, etc., and those skilled in the art can flexibly and flexibly implement it. The deep learning algorithm can be a logistic regression model suitable for binary classification, or it can be a twin tower model, and those skilled in the art can flexibly and flexibly implement it.

[0104] Step S1502: for various recommendation scenarios associated with new independent station stores, obtain practical recommendation strategies and their measured recommendation effectiveness scores used by the multiple independent station stores to recommend products in the recommendation scenarios, screen out the practical recommendation strategy with the highest measured recommendation effectiveness, and set it as the practical recommendation strategy used by the new independent station stores to recommend products in the recommendation scenarios.

[0105] Similarly, in a newly-born independent station store, recommendation scenarios are pre-configured for various product recommendation events that occur therein. The recommendation scenarios may refer to the recommendation of pages and / or behaviors that trigger corresponding product recommendation events, or may refer to the recommendation of activities formulated based on the operation and / or marketing conditions of the independent station store.

[0106] Taking a recommendation scenario associated with a new independent station store as an example, obtain the practical recommendation strategies and their measured recommendation effectiveness scores determined in previous iterations by multiple independent station stores corresponding to the same type of stores as the new independent station store, and screen out the practical recommendation strategy with the highest measured recommendation effectiveness, which is the most suitable recommendation strategy for the new independent station store to recommend products in the recommendation scenario. Therefore, the screened practical recommendation strategy is set as the practical recommendation strategy used by the new independent station store to recommend products in the recommendation scenario.

[0107] In this embodiment, by determining multiple independent station stores of the same type as the new independent station store, and then for various recommendation scenarios associated with the new independent station store, the practical recommendation strategy with the best actual recommendation effect can be selected from the practical recommendation strategies of the multiple independent station stores of the same type, and set as the practical recommendation strategy used in the recommendation scenario corresponding to the new independent station store. It can be seen that the cold start problem of the new independent station store caused by insufficient user volume can be solved, and recommendation strategies corresponding to various recommendation scenarios can be set for the new independent station store.

[0108] See also Figure 3 , a recommendation strategy dynamic adjustment device provided to adapt to one of the purposes of the present application is a functional embodiment of the recommendation strategy dynamic adjustment method of the present application. On the other hand, the device is a recommendation strategy dynamic adjustment device provided to adapt to one of the purposes of the present application, including a collection acquisition module 1100, an iteration start module 1200, an iteration continuation module 1300 and an iteration end module 1400, wherein the collection acquisition module 1100 is used to determine the target recommendation scenario from various recommendation scenarios associated with the independent station store, and obtain a recommendation strategy set constructed corresponding to the target recommendation scenario based on the recommendation stage strategy library, wherein the recommendation strategy set includes multiple recommendation strategies; the iteration start module 1200 is used to start the iteration process, and the online application recommendation After each recommendation strategy in the strategy set, the recommendation strategy with the highest measured recommendation effectiveness score is determined as the preferred recommendation strategy; the iteration continuation module 1300 is used to reconstruct a new recommendation strategy for the target recommendation scenario based on the recommendation stage strategy library, and replace all recommendation strategies other than the preferred recommendation strategy in the recommendation strategy set with the new recommendation strategy, and continue the iteration process, wherein the predicted recommendation effectiveness score of the new recommendation strategy is higher than the measured recommendation effectiveness score of the preferred recommendation strategy; the iteration end module 1400 is used to end the iteration process when the new recommendation strategy cannot be constructed, and obtain the latest preferred recommendation strategy for the target recommendation scenario as a practical recommendation strategy for product recommendation in the target recommendation scenario.

[0109] In a further embodiment, the iterative startup module 1200 includes: a measured effectiveness sub-module, which is used to call each recommendation strategy in the recommendation strategy set to respectively respond to at least one online recommendation request associated with the target recommendation scenario, and determine the measured recommendation effectiveness score of each recommendation strategy; a strategy two-selection sub-module, which is used to call the two recommendation strategies with higher measured recommendation scores to respectively respond to at least one online recommendation request associated with the recommendation scenario, determine the new measured recommendation effectiveness scores corresponding to the two recommendation strategies, and update the original measured recommendation scores; a strategy single-selection sub-module, which is used to determine that the preferred recommendation strategy is the recommendation strategy with the highest measured recommendation effectiveness score among the two recommendation strategies.

[0110] In a further embodiment, the iterative continuation module 1300 includes: a candidate construction submodule, which is used to reconstruct at least one candidate recommendation strategy corresponding to the recommendation scenario based on at least one stage strategy corresponding to the recommendation strategies set for each stage in the recommendation process in the recommendation stage strategy library; an effectiveness prediction submodule, which is used to predict the recommendation effectiveness score of each candidate recommendation strategy using a preset effectiveness prediction model; and a new strategy confirmation submodule, which is used to confirm that the candidate recommendation strategy is a new recommendation strategy when the predicted recommendation effectiveness score of the candidate recommendation strategy is higher than the actual recommended effectiveness score of the latest preferred recommendation strategy.

[0111] In a further embodiment, the effectiveness prediction submodule includes: a training preparation submodule for obtaining training samples and supervisory labels in a prepared training set, wherein the training samples include historical recommendation strategies and applicable recommendation scenarios, and the supervisory labels are the measured recommendation effectiveness scores corresponding to the product recommendations of the historical recommendation strategies for the recommendation scenarios; a feature extraction submodule for inputting the training samples into the effectiveness prediction model to extract corresponding deep semantic features and obtain an effectiveness feature vector; a classification mapping submodule for classifying and mapping the effectiveness feature vector to predict the corresponding predicted recommendation effectiveness score; a model tuning submodule for calculating the loss value corresponding to the predicted recommendation effectiveness score based on the supervisory labels, and updating the effectiveness prediction model based on the loss value until the model converges, so as to predict the recommendation effectiveness score corresponding to the product recommendations of the recommendation strategies for their applicable recommendation scenarios.

[0112] In a further embodiment, after the iteration end module 1400, it includes: an event accumulation sub-module, used to determine the total number of events corresponding to the difference in the recommended effectiveness score of the practical recommendation strategy that decreases by more than a first preset threshold; a strategy update sub-module, used to restart the iteration process when the total number exceeds a second preset threshold to redetermine a new practical recommendation strategy.

[0113] In a further embodiment, the iterative startup module 1200 includes: a distribution initialization submodule, which is used for obtaining the actual recommendation effectiveness score of each recommendation strategy in the recommendation strategy set if the recommendation strategy is a preferred recommendation strategy, and obtaining the predicted recommendation effectiveness score of the recommendation strategy if the recommendation strategy is not a preferred recommendation strategy, and constructing the effectiveness distribution of each recommendation strategy according to the corresponding initialization of the recommendation effectiveness score of each recommendation strategy, wherein the predicted recommendation effectiveness score is obtained for the corresponding recommendation strategy using a preset effectiveness prediction model; a distribution update submodule, which is used to start the sampling iteration process, and randomly sample each of the effectiveness distributions to determine Corresponding to the expected probability of effectiveness, after calling the recommendation strategy with the highest expected probability of effectiveness and applying it online to multiple online recommendation requests associated with the target recommendation scenario, the total number of valid recommendations and the total number of invalid recommendations actually measured by the recommendation strategy are determined accordingly, and the effectiveness distribution of the recommendation strategy is updated according to the total number of valid recommendations and the total number of invalid recommendations, and the sampling iteration process continues; the strategy optimization submodule is used to end the sampling iteration process when the number of sampling iterations exceeds a preset threshold, determine the actually measured recommendation effectiveness score of each recommendation strategy according to the latest effectiveness distribution of each recommendation strategy, and determine the recommendation strategy with the highest actually measured recommendation effectiveness score as the preferred recommendation strategy.

[0114] In a further embodiment, after the iteration end module 1400, it includes: an event response submodule, which is used to respond to the recommendation strategy setting event of the new independent station store, and determine that the new independent station store associated with the event belongs to multiple independent station stores corresponding to the same type of stores; a strategy setting submodule, which is used to obtain the practical recommendation strategies and their actual measured recommendation effectiveness scores used by the multiple independent station stores to recommend products in the recommendation scenarios associated with the new independent station stores, and screen out the practical recommendation strategies with the highest measured recommendation effectiveness, and set them as the practical recommendation strategies used by the new independent station stores to recommend products in the recommendation scenarios.

[0115] In order to solve the above technical problems, the present application also provides a computer device. Figure 4As 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 method for dynamically adjusting the recommendation strategy. 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 method for dynamically adjusting the recommendation strategy 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.

[0116] In this embodiment, the processor is used to execute Figure 3 The 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 recommended strategy dynamic adjustment device of this application, and the server can call the program code and data of the server to execute the functions of all submodules.

[0117] 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 method for dynamically adjusting the recommendation strategy of any embodiment of the present application.

[0118] 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).

[0119] In summary, this application can fully explore the optimal recommendation strategy for the scenario and is expected to achieve the expected recommendation results.

[0120] 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.

[0121] 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 method for dynamically adjusting a recommendation strategy, characterized in that: The steps include: Determine a target recommendation scenario from various recommendation scenarios associated with the independent station store, and obtain a recommendation strategy set corresponding to the target recommendation scenario based on the recommendation stage strategy library, wherein the recommendation strategy set includes multiple recommendation strategies; Start the iterative process, apply each recommendation strategy in the recommendation strategy set online, and then determine the recommendation strategy with the highest measured recommendation effectiveness score as the preferred recommendation strategy; Reconstructing a new recommendation strategy for the target recommendation scenario based on the recommendation phase strategy library, replacing all recommendation strategies other than the preferred recommendation strategy in the recommendation strategy set with the new recommendation strategy, and continuing the iterative process, wherein the predicted recommendation effectiveness score of the new recommendation strategy is higher than the measured recommendation effectiveness score of the preferred recommendation strategy; When the new recommendation strategy cannot be constructed, the iterative process is terminated, and the latest preferred recommendation strategy for the target recommendation scenario is obtained as a practical recommendation strategy for product recommendations in the target recommendation scenario.

2. The method for dynamically adjusting the recommendation strategy according to claim 1, characterized in that: After applying each recommendation strategy in the recommendation strategy set online, the recommendation strategy with the highest measured recommendation effectiveness score is determined as the preferred recommendation strategy, including the following steps: Invoke each recommendation strategy in the recommendation strategy set to respectively respond to at least one online recommendation request associated with the target recommendation scenario, and determine a measured recommendation effectiveness score of each recommendation strategy; Invoke the two recommendation strategies with higher measured recommendation scores to respectively respond to at least one online recommendation request associated with the recommendation scenario, determine new measured recommendation effectiveness scores corresponding to the two recommendation strategies, and update the original measured recommendation scores; The preferred recommendation strategy is determined to be the recommendation strategy with the highest measured recommendation effectiveness score among the two recommendation strategies.

3. The method for dynamically adjusting the recommendation strategy according to claim 1, characterized in that: Reconstructing a new recommendation strategy for the target recommendation scenario based on the recommendation stage strategy library includes the following steps: Reconstructing at least one candidate recommendation strategy applicable to the recommendation scenario based on at least one corresponding stage strategy among the recommendation strategies set for each stage in the recommendation process in the recommendation stage strategy library; Use the preset effectiveness estimation model to predict the recommendation effectiveness score of each candidate recommendation strategy; When the predicted recommendation effectiveness score of the candidate recommendation strategy is higher than the actual measured recommendation effectiveness score of the latest preferred recommendation strategy, the candidate recommendation strategy is confirmed as a new recommendation strategy.

4. The method for dynamically adjusting the recommendation strategy according to claim 3, characterized in that: Before using the preset effectiveness estimation model to predict the recommendation effectiveness score of each candidate recommendation strategy, the following steps are included: Obtain training samples and supervisory labels in a prepared training set, wherein the training samples include historical recommendation strategies and applicable recommendation scenarios, and the supervisory labels are actually measured recommendation effectiveness scores corresponding to the commodity recommendations of the historical recommendation strategies in the recommendation scenarios; Inputting the training sample into the effectiveness prediction model to extract corresponding deep semantic features and obtain an effectiveness feature vector; Classify and map the effectiveness feature vector to predict the corresponding prediction recommendation effectiveness score; The loss value corresponding to the predicted recommendation effectiveness score is calculated according to the supervision label, and the effectiveness estimation model is updated according to the loss value until the model converges, so as to predict the recommendation effectiveness score corresponding to the product recommendation of the recommendation strategy for its applicable recommendation scenario.

5. The method for dynamically adjusting the recommendation strategy according to claim 1, characterized in that: After obtaining the latest preferred recommendation strategy for the target recommendation scenario as a practical recommendation strategy, the following steps are included: Determine the total number of events corresponding to the decrease in the difference of the recommendation effectiveness score actually measured by the practical recommendation strategy exceeding the first preset threshold; When the total number of times exceeds a second preset threshold, the iterative process is restarted to redetermine a new practical recommendation strategy.

6. The method for dynamically adjusting the recommendation strategy according to claim 1, characterized in that: After applying each recommendation strategy in the recommendation strategy set online, the recommendation strategy with the highest measured recommendation effectiveness score is determined as the preferred recommendation strategy, including the following steps: For each recommendation strategy in the recommendation strategy set, if the recommendation strategy is a preferred recommendation strategy, the measured recommendation effectiveness score of the recommendation strategy is obtained; if the recommendation strategy is not a preferred recommendation strategy, the predicted recommendation effectiveness score of the recommendation strategy is obtained; and the effectiveness distribution of each recommendation strategy is constructed based on the corresponding initialization of the recommendation effectiveness score of each recommendation strategy, wherein the predicted recommendation effectiveness score is obtained for the corresponding recommendation strategy using a preset effectiveness prediction model; Initiate a sampling iteration process, randomly sample each of the effectiveness distributions to determine the corresponding expected effectiveness probability, call the recommendation strategy with the highest expected effectiveness probability and apply it online to multiple online recommendation requests associated with the target recommendation scenario, and then determine the total number of valid recommendations and the total number of invalid recommendations measured by the recommendation strategy accordingly, update the effectiveness distribution of the recommendation strategy according to the total number of valid recommendations and the total number of invalid recommendations, and continue the sampling iteration process; When the number of sampling iterations exceeds the preset threshold, the sampling iteration process is terminated, and the measured recommendation effectiveness scores of each recommendation strategy are determined according to the latest effectiveness distribution of each recommendation strategy, and the recommendation strategy with the highest measured recommendation effectiveness score is determined as the preferred recommendation strategy.

7. The method for dynamically adjusting the recommendation strategy according to any one of claims 1 to 6, characterized in that: After obtaining the latest preferred recommendation strategy for the target recommendation scenario as a practical recommendation strategy, the following steps are included: In response to a recommendation strategy setting event of a new independent station store, it is determined that the new independent station store associated with the event belongs to multiple independent station stores corresponding to the same type of store; For various recommendation scenarios associated with new independent station stores, obtain the practical recommendation strategies used by the multiple independent station stores to recommend products in the recommendation scenarios and their actually measured recommendation effectiveness scores, screen out the practical recommendation strategy with the highest measured recommendation effectiveness, and set it as the practical recommendation strategy used by the new independent station stores to recommend products in the recommendation scenarios.

8. A device for dynamically adjusting a recommendation strategy, characterized in that: include: A set acquisition module, used to determine a target recommendation scenario from various recommendation scenarios associated with independent station stores, and acquire a recommendation strategy set corresponding to the target recommendation scenario based on a recommendation stage strategy library, wherein the recommendation strategy set includes multiple recommendation strategies; The iteration start module is used to start the iteration process. After applying each recommendation strategy in the recommendation strategy set online, the recommendation strategy with the highest recommendation effectiveness score is determined as the preferred recommendation strategy. An iterative continuation module is used to reconstruct a new recommendation strategy for the target recommendation scenario based on the recommendation phase strategy library, replace all recommendation strategies other than the preferred recommendation strategy in the recommendation strategy set with the new recommendation strategy, and continue the iterative process, wherein the predicted recommendation effectiveness score of the new recommendation strategy is higher than the measured recommendation effectiveness score of the preferred recommendation strategy; The iteration ending module is used to end the iteration process when the new recommendation strategy cannot be constructed, and obtain the latest preferred recommendation strategy of the target recommendation scenario as a practical recommendation strategy for product recommendation in the target recommendation scenario.

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