Agent-based recommendation method, apparatus, computing device, medium, and program

Abstract

The application provides an agent-based recommendation method, device, computing equipment, medium and program, belonging to the field of electronic commerce. The agent-based recommendation method comprises: acquiring at least one demand intention input by a user; generating a decision variable based on a plurality of products to be recommended, the decision variable comprising a decision component for each product to be recommended; generating a recommendation constraint condition according to a recommendation condition of the plurality of products to be recommended; generating a recommendation target function based on the at least one demand intention; solving the decision variable according to the recommendation target function and the recommendation constraint condition to determine a target product from the plurality of products to be recommended; and recommending the target product to the user. The method can accurately understand the user's intention, generate a complete recommendation scheme, support cross-granularity recommendation, and improve the recommendation effect through multi-objective optimization.

Description

Technical Field

[0001] This application relates to the field of e-commerce, and more particularly to a recommendation method, apparatus, computing device, medium, and program based on intelligent agents. Background Technology

[0002] Currently, with the rapid development of artificial intelligence (AI) technology, in order to better serve users, major companies are providing intelligent recommendation services, suggesting products or services that users may be interested in. Therefore, ensuring that the recommended content better matches user needs has become particularly important.

[0003] Existing recommendation methods often fail to adequately meet users' subjective desires and needs during consumption. They cannot accurately identify and analyze user needs and recommendation preferences, lack personalized recommendation capabilities, and the recommended results often fail to satisfy user needs. Specifically, traditional recommendation algorithms lack interaction with user intent, cannot accurately analyze customers' true needs, and typically only recommend individual items, lacking an understanding and fulfillment of users' explicit needs. This "one-sided" recommendation approach may not match the user's current shopping needs, affecting user experience and recommendation effectiveness. Secondly, traditional recommendation results are usually a sorted list of individual items rather than a complete shopping plan. Users still need to manually select and combine items from the recommended list, failing to directly obtain a complete shopping plan that meets requirements such as the number of diners, tastes, and budget. Furthermore... Existing intelligent agent systems typically only perform recommendations at a high-level granularity, failing to automatically handle sub-item recommendations or cross-granularity combinations. For example, they can only recommend at the package level, and cannot perform sub-item-level recommendations or combinations between sub-items and packages when multiple sub-items exist within a package. Furthermore, existing recommendation systems usually only consider the rating of the recommendation results, ranking recommended products based on the rating, lacking comprehensive consideration of other objectives such as user intent error, efficiency and cost, and diversity balance. Therefore, they cannot simultaneously meet user needs while ensuring recommendation quality, cost, and richness. Thus, there is an urgent need for a recommendation method that can accurately identify user recommendation needs, combine explicit user requirements with personalized recommendations, provide complete shopping solutions, and optimize across product granularities. Summary of the Invention

[0004] To address the aforementioned issues, there is an urgent need for a recommendation method that can accurately understand user intent, generate complete recommendation schemes, support cross-granularity recommendations, and improve recommendation performance through multi-objective optimization. This application aims to provide an agent-based recommendation method that models user intent as mathematical constraints and objective functions, thereby enabling precise identification and analysis of user needs. Based on these needs, it determines the products or services to recommend to the user, optimizes comprehensive objectives such as matching degree, cost, and diversity of recommendation results, and supports the complete assembly and output of product structures, thus improving recommendation results and better meeting users' shopping intent needs.

[0005] An embodiment of the first aspect of this application provides an agent-based recommendation method, comprising: acquiring at least one demand intent input by a user; generating decision variables based on multiple products to be recommended, the decision variables including decision components for each product to be recommended; generating recommendation constraints based on recommendation conditions for the multiple products to be recommended; generating a recommendation objective function based on at least one demand intent; solving the decision variables according to the recommendation objective function and the recommendation constraints to determine a target product from the multiple products to be recommended; and recommending the target product to the user.

[0006] An embodiment of the second aspect of this application provides an agent-based recommendation device, comprising: an acquisition module for acquiring at least one demand intent input by a user; a first generation module for generating decision variables based on a plurality of products to be recommended, the decision variables including decision components for each product to be recommended; a second generation module for generating recommendation constraints based on recommendation conditions for the plurality of products to be recommended; a third generation module for generating a recommendation objective function based on at least one demand intent; a solving module for solving the decision variables based on the recommendation objective function and the recommendation constraints to determine a target product from the plurality of products to be recommended; and a recommendation module for recommending the target product to the user.

[0007] An embodiment of the third aspect of this application provides a computing device, including: at least one processor; and at least one memory communicatively connected to the at least one processor, the at least one memory storing instructions that, when executed individually or jointly by the at least one processor, cause the computing device to perform the recommended method in the above embodiments.

[0008] An embodiment of the fourth aspect of this application provides a computer-readable storage medium storing instructions that, when executed individually or jointly by one or more processors of a computing device, cause the computing device to perform the recommended method described above.

[0009] An embodiment of the fifth aspect of this application provides a computer program product including instructions that, when executed individually or jointly by one or more processors of a computing device, cause the computing device to perform the recommended method described above.

[0010] In the technical solution of this application embodiment, by parsing user input into structured intent parameters and achieving accurate matching through mathematical modeling, the user's needs can be accurately identified and met, realizing the combination of user needs and personalized recommendations. Using the recommendation method in this application embodiment, a complete shopping plan can be provided, enabling not only single-item recommendations but also package combinations and sub-items within packages, forming an end-to-end complete recommendation solution. It has cross-granularity recommendation capabilities, supporting combined recommendations of single items, packages, and sub-items, effectively improving the flexibility of recommendations. During the recommendation process, various objectives such as intent error, recommendation score, efficiency cost, and diversity are comprehensively considered to achieve comprehensive optimization of the recommendation effect. The algorithm automatically completes processes such as intent parsing, objective function construction, optimization solution, and solution assembly, effectively improving the intelligence and efficiency of the recommendation process.

[0011] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0012] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments disclosed in this application and should not be construed as limiting the scope of this application.

[0013] Figure 1 This is a flowchart illustrating some embodiments of the agent-based recommendation method of this application; Figure 2 This is a flowchart illustrating the process of generating the recommendation objective function in some embodiments of this application; Figure 3 This is a flowchart illustrating the process of generating the recommendation objective function in some embodiments of this application; Figure 4 This is a flowchart illustrating the weighted acquisition of the recommended objective function in some embodiments of this application; Figure 5 This is a flowchart illustrating the process of determining the recommended objective function in some embodiments of this application; Figure 6 This is a schematic diagram of the process for determining the target product in some embodiments of this application; Figure 7 This is a flowchart illustrating the process of handling demand intent in some embodiments of this application; Figure 8 This is a schematic block diagram of an agent-based recommendation device according to some embodiments of this application; Figure 9 This is a schematic block diagram of a computing device according to some embodiments of this application; Figure 10 This is a flowchart illustrating an agent-based recommendation method according to some embodiments of this application. Detailed Implementation

[0014] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0015] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0016] In the description of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In this application, "multiple" means two or more, and "multi-level" refers to two or more levels, unless otherwise explicitly defined.

[0017] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0018] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0019] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), and similarly, "multiple groups" refers to two or more (including two groups).

[0020] In the technical solution of this application, the collection, storage, use, processing, transmission, provision and disclosure of user personal information are all done with the consent of the user, comply with the provisions of relevant laws and regulations, and do not violate public order and good morals.

[0021] Currently, with the development of artificial intelligence (AI) technology, in order to better serve users, major companies are providing intelligent recommendation services to recommend products or services that users may be interested in.

[0022] Existing recommendation methods mainly include using traditional recommendation algorithms to rank products and recommend them to users according to the ranking, or using intelligent agents to filter products based on users' shopping needs. These methods typically fail to adequately meet users' subjective desires and needs during consumption, lack personalized recommendation capabilities, and the recommended results do not fully satisfy user requirements.

[0023] To better meet user needs and achieve intelligent recommendations based on those needs, an agent-based recommendation method can be used. This method involves obtaining the user's input intent into the agent, determining and solving the corresponding decision variables, constraints, and objective function to ultimately obtain the recommendation result. Recommendations obtained using this method can achieve personalized intelligent recommendations while meeting the user's explicitly expressed intent, thus significantly improving the user experience.

[0024] This application provides an agent-based recommendation method. (Reference) Figure 1 The agent-based recommendation method 100 includes steps 110 to 160.

[0025] Step 110: Obtain at least one user input request intent.

[0026] Step 120: Generate decision variables based on multiple products to be recommended. Decision variables include decision components for each product to be recommended.

[0027] Step 130: Generate recommendation constraints based on the recommendation conditions of multiple products to be recommended.

[0028] Step 140: Generate a recommendation objective function based on at least one demand intent.

[0029] Step 150: Solve for the decision variables based on the recommendation objective function and recommendation constraints to determine the target product from multiple products to be recommended.

[0030] Step 160: Recommend the target product to the user.

[0031] In the embodiments of this application, the "product to be recommended" may include different types of products that can be provided to users. Depending on the application scenario, the product to be recommended may be a physical product, a virtual product, or a service product. The following description uses a food ordering scenario as an example, where the product to be recommended may be different types of dishes on a menu. The "target product" is the product selected from all the products to be recommended and recommended to the user. It should be understood that, based on the solution results, the target product may be a single product or multiple products; this application does not limit the specific number of target products.

[0032] Each demand intent can represent a user's shopping demand for goods or services. In step 110, the user's demand intent will be obtained based on user input. In one example, the user can input shopping demands into the agent, and the agent will obtain the user's demand intent based on the user's input. In some embodiments, the demand intent can be represented in the form of intent parameters. Intent parameters are a formatted parameter language that is universally applicable to various scenarios for expressing user shopping demands. They can be generated in different ways, such as by a large language model converting the user's input natural language into the form of intent parameters, or by the user filling in intent parameters through corresponding operations on the operation interface. Intent parameters include a combination of product range, product metrics and / or symbols and thresholds. In one example, each intent parameter can be composed of product range (e.g., the product code (ID), name, label, etc.), product metrics (e.g., product quantity, number of people suitable, total product price, etc.), symbols (e.g., =, !=, >, >=, <, <=, etc.), and thresholds (i.e., specific numerical values), such as ID=12345, flavor != spicy, category = beverage and temperature != ice, etc.

[0033] According to some embodiments of this application, at least one demand intent includes at least one of the following: user demand quantity, recommended product range demand, product demand quantity, product price demand, and user behavior preference.

[0034] When users input their shopping needs, they may involve various aspects. These include: the quantity of products a user desires can be the number of people the product is suitable for; the range of recommended products can include product categories (e.g., product names, codes, labels, etc.); special product requirements (e.g., no ice, no spice); the quantity of products required; in the example, this quantity requirement can be related to the user's desired quantity or entered separately by the user; the price of products can include the expected price for certain products; and user behavior preferences can include the user's preferred product preferences.

[0035] Taking the food ordering scenario as an example, a user's shopping needs may include factors such as the number of diners, the type, quantity, and price of the food they wish to order.

[0036] In one example, the user inputs the following language into the agent: "Five people dining together, no spicy food, total beverage price not exceeding 100, and one A-activity set meal." In step 110, the user's intent will be obtained based on this input and converted into intent parameters. For example, "five people dining together" can be converted into the intent parameter "suitable number of people = 5", "no spicy food" can be converted into the intent parameter "product quantity (flavor != spicy) = 0" (i.e., the quantity of products with a non-spicy flavor is 0), "total beverage price not exceeding 100" can be converted into the intent parameter "total product price (category = beverage) <= 100" (i.e., the total price of beverage products is less than or equal to 100), and "one A-activity set meal" can be converted into the intent parameter "product quantity (product type = set meal and activity = A) = 1" (i.e., the quantity of the product type set meal in activity A is 1).

[0037] According to some embodiments of this application, the recommended method 100 further includes step 170. (See reference...) Figure 7 Step 170 includes at least one of steps 710 to 730.

[0038] Step 710: Determine the number of products requested by the user based on the number of user requests and the user's historical intent data.

[0039] Step 720: Update the range of recommended products entered by the user based on the product parameters of multiple products to be recommended.

[0040] Step 730: Based on at least one requirement intent, determine the parameter range corresponding to any one of the at least one requirement intents.

[0041] Step 170 can be understood as a step of verifying and rewriting the demand intent obtained in step 110. In step 710, as mentioned above, the product quantity demand can be related to the user demand quantity. Therefore, the user's input user demand quantity can be converted into the product demand quantity, making it easier to determine the target product from the products to be recommended. Taking "Applicable number of people = 5" as an example, based on the user's historical intent data, that is, the user's demand intent data in the historical process, it is converted into the required product quantity. For example, if the user usually orders staple food, snacks, and beverages in the historical ordering process, then "Applicable number of people = 5" can be rewritten as "Product quantity (category = staple food) = 5, Product quantity (category = snacks) = 5, Product quantity (category = beverages) = 5", that is, the product demand quantity of the main category is 5, the product demand quantity of the snack category is 5, and the product demand quantity of the beverage category is 5.

[0042] In step 720, the user's input intent can be corrected and updated, i.e., intent rewriting. If no product fully matches a particular intent after evaluation, the intent will be corrected based on the actual situation, retaining the range of recommended products with a higher probability of matching. Specifically, in step 720, product parameters for each product to be recommended can be obtained, representing the attribute information of each product. Based on the user's input range of recommended products and the product parameters of each product to be recommended, if none of the recommended products fully match the user's input range, the user's input range of recommended products will be updated, i.e., rewritten, to better reflect the actual product parameters, thus selecting the corresponding product to be recommended. In one example, the user inputs "I want a three-piece set meal of X", but there is no three-piece set meal named X among the recommended products. A query for X shows that the number of sub-items within the set meal is 4. This means the intent parameter is represented as "Product Quantity (Product Name = X Three-Piece Set meal and Number of Sub-items in Set Meal = 4) = 1". However, X is not a three-piece set but a four-piece set, creating a contradiction between the intent parameter "X Three-Piece Set meal" and "Number of Sub-items in Set Meal = 4". In this case, the intent parameter will be modified based on the actual situation, retaining the recommended product range with a higher probability of being desired. For example, "X Three-Piece Set meal" might be updated to "X Four-Piece Set meal" to match the actual product parameters.

[0043] In step 730, to facilitate subsequent calculations, the parameter range corresponding to each demand intent can be determined. For example, if the intent parameter is expressed as > a certain threshold or < a certain threshold, i.e., an open interval, it can be redefined as >= a certain threshold or <= a certain threshold, i.e., a closed interval. As mentioned above, the symbols involved in the intent parameters can include =, !=, >, >=, <, <=, etc. By redefining the parameter range and converting the open interval into a closed interval, the symbols ">" and "<" can be omitted in subsequent calculations, facilitating subsequent processing. In one example, > threshold or < threshold can be redefined as >= threshold + minimum change or <= threshold - minimum change, respectively. The minimum change for product quantity can be 1, and the minimum change for product price can be 0.01. For example, if the user's intent parameter is "total product price (category = beverage) < 100", it can be modified to "total product price (category = beverage) <= 100 - 0.01"; if the user's intent parameter is "product quantity (category = beverage) > 5", it can be modified to "product quantity (category = beverage) >= 5 + 1".

[0044] By rewriting and updating the obtained demand intent, the demand intent can be better matched with the products to be recommended, thereby obtaining more accurate target products to recommend to users.

[0045] In step 120, a decision variable will be generated for all products to be recommended. Each product to be recommended will correspond to one component of the decision variable. By solving for this decision variable, the final recommendation result can be determined.

[0046] According to some embodiments of this application, the multiple products to be recommended include n products, where n is a natural number greater than 1, and the decision variables are constructed as follows: , in, As decision variables, Each of the n products to be recommended corresponds to one of them, and indicates the number of recommended products for each product.

[0047] If there are a total of n types of products available for purchase, then the decision variable can be the quantity of each type of product to be recommended.

[0048] Continuing with the example of ordering food, the recommended products are all purchasable items in the menu, including all individual items and all set meals (and any possible sub-items included in a set meal). There are n types of these products. For each product, there is a corresponding decision component in the decision variables, i.e. The recommendation process involves determining the value of each decision component, i.e., the number of recommendations for each type of product.

[0049] In step 130, recommendation constraints will be generated based on the different recommendation conditions of the products to be recommended. That is, when determining each decision component in the decision variables, these recommendation constraints must be met. Recommendation conditions indicate the conditions that the products to be recommended must meet when making a recommendation, and can be determined based on factors such as the type of product or its usage scenario. Recommendation constraints can be flexibly determined from different aspects depending on the usage scenario. In some embodiments, recommendation constraints may include one or more of the following: constraints related to package structure, constraints related to discount conflicts, constraints related to product purchase limits, constraints related to add-to-cart recommendations, and constraints related to sales strategies.

[0050] According to some embodiments of this application, the recommended constraints are constructed as any one of the following: ， ， ， in, Let j be the coefficient vector corresponding to the j-th constraint. Let be the constraint value that the j-th constraint needs to satisfy.

[0051] Depending on the different recommendation conditions, one of the three forms mentioned above can be selected to obtain the recommendation constraints.

[0052] In one example, the recommendation condition could be a package structure condition. For instance, for a certain package (ID 1), there are 5 sub-items to choose from (assuming each sub-item has IDs 2, 3, 4, 5, and 6), and 3 of these sub-items must be selected before the package can be recommended. Under this recommendation condition, the recommendation constraint can be in the first form. Specifically, it can be expressed as: 。

[0053] in These are the decision components for the package (ID 1) and the five sub-items (IDs 2, 3, 4, 5, 6), respectively. The above formula can be understood as follows: the package quantity will only be 1 if 3 sub-items are selected.

[0054] Simplifying the above expression, it can be represented as: 。

[0055] Right now = ， 。

[0056] In another example, the recommendation condition could be a coupon usage recommendation condition, such as multiple coupons cannot be used simultaneously, and only one can be used at most. It requires determining from among the multiple usable coupons whether to recommend one coupon to the user or not use any coupon; that is, only one coupon can be used (coupon quantity is 1) or no coupon can be used (coupon quantity is 0). Under this recommendation condition, the recommendation constraint can choose the second form. That is, the sum of the decision components corresponding to all coupons must be less than or equal to 1.

[0057] In another example, the recommendation condition could be a product purchase limit, such as a maximum purchase of one unit of a certain product. Under this recommendation condition, the recommendation constraint can choose the second form. That is, the decision component corresponding to the product needs to be less than or equal to 1.

[0058] In another example, the recommendation condition could be an add-to-cart recommendation condition. For instance, when a user adds items to their cart after their initial order, it's necessary to ensure that the unmodified portions remain unchanged; that is, the recommendation result should retain at least the current quantity of the unmodified portions. Under this recommendation condition, the recommendation constraint can choose a third form. That is, the decision components corresponding to these unmodified products need to be greater than or equal to the current quantity.

[0059] It should be understood that the above recommended conditions are merely examples. Depending on the specific use case, any of the three recommended constraints can be chosen. and Both can be configured. For example, products B and C may need to be sold together, meaning that the decision components for products B and C must both be either 0 or both be 1. This application does not limit the specific formula for the recommended constraints.

[0060] By defining reasonable constraints, the final recommendation results can meet product constraints, enabling users to make subsequent purchases and use based on the recommendations, thus reducing the probability of being unable to perform subsequent operations.

[0061] In step 140, an objective function for the recommendation process can be generated based on the user's needs and intentions. Under the constraints of the recommendation conditions, the decision variables are solved based on the objective function.

[0062] In determining the recommendation objective function, it can be generated from different aspects.

[0063] According to some embodiments of this application, reference is made to Figure 2 Step 140 includes steps 210 to 220.

[0064] Step 210: Use at least one recommendation model to determine the recommendation score corresponding to the decision variable, so as to generate recommendation score functions corresponding to at least one recommendation model respectively.

[0065] Step 220: Generate the recommendation objective function based at least on the recommendation scoring function.

[0066] When making recommendations, several recommendation models can be used to score the recommendation results from different perspectives, resulting in corresponding recommendation scoring functions. For example, recommendation models can score the recommendation results corresponding to decision variables based on aspects such as "the degree of matching with the user's personal interests," "the popularity of the recommendation results," and "the popularity of the activities involved in the recommendation results."

[0067] In step 210, it is assumed that multiple recommendation models are used for the decision variables respectively. Recommendations are rated. Each recommendation model will output a separate rating vector. ,in Let represent the y-th recommendation model, where each component of the rating vector represents a rating for each product to be recommended. The dimension of the rating vector is the same as the dimension of the decision variables. The recommendation rating function corresponding to the y-th recommendation model is... It can be represented as: 。

[0068] In step 220, the recommendation objective function can be obtained solely from the recommendation scoring function. In one example, the recommendation objective function can be expressed as maximizing the sum of the recommendation scoring functions obtained from all recommendation models. Based on this objective function, corresponding target products can be recommended to the user.

[0069] In other embodiments, when determining the recommendation objective function, in addition to using a recommendation model for scoring, other optimization objectives may be considered.

[0070] According to some embodiments of this application, reference is made to Figure 3 Step 140 also includes steps 211 to 213.

[0071] Step 211: Generate the intent error function corresponding to the decision variable based on at least one demand intent.

[0072] Step 212: Generate the efficiency cost function corresponding to the decision variables based on the number of recommendations for each product to be recommended.

[0073] Step 213: Generate the group quantity function corresponding to the decision variable based on the product range of each product to be recommended.

[0074] To achieve better recommendation results, factors such as the error between the recommended results and the user's intended needs, as well as the number and variety of products in the recommended results, can be considered. The intent error function measures the deviation between the recommended product combination and the user's intended needs; it can also be understood as the difference between the identified target products and the user's actual needs. The efficiency cost function limits the number of identified target products, preventing the recommendation of redundant products to improve the recommendation score. The group quantity function increases the diversity of recommended products, preventing the presence of too many duplicate products in the target product pool.

[0075] In step 211, the error between the recommendation result and the user's intended needs will be determined, that is, the intention error function will be generated.

[0076] According to some embodiments of this application, the intent error function is constructed as one of the following: ， ， ， in, Let i be the intent error function corresponding to the i-th demand intent. It is an absolute value function. , Let i be the coefficient vector corresponding to the i-th demand intention. This represents the demand value corresponding to the i-th demand intent.

[0077] As mentioned above, after step 170, the symbols involved in the intent parameters can include =, >=, <=, etc. The three intent error functions mentioned above can correspond to the three symbols =, >=, and <=, respectively.

[0078] According to some embodiments of this application, the coefficient vector corresponding to the i-th demand intent is... The components in can be constructed as: ， in, For the p-th decision component in the decision variables The corresponding coefficients, Represents the p-th decision component Does it satisfy the i-th requirement? Represents the p-th decision component The measure of the i-th demand intention.

[0079] In one example, if the p-th decision component If the i-th demand intention is satisfied, then =1, if the p-th decision component If the i-th requirement intention is not satisfied, then =0.

[0080] The value of can be set according to the usage scenario. In the example, if the metric in the i-th demand intent is the quantity of products, then the metric value of each product to be recommended can be 1; if the metric in the i-th demand intent is the total price of the goods, then the metric value of each product to be recommended is the price of that product.

[0081] Taking the i-th demand intent as "total product price (category = beverage) <= 100" as an example, that is, requiring the total price of products of the beverage category to be less than or equal to 100, in this example, The value is the price of the product; if this intention is met, then... =1.

[0082] By determining the error between the recommended results and the user's intended needs, we can accurately reflect whether the recommended results adequately meet the user's needs. In practical applications, the smaller the intent error function, the better the results match the user's needs.

[0083] While satisfying user intent, and considering the recommendation model's score, it's easy for the number of recommended products to correlate with the recommendation score, potentially resulting in an infinite number of recommended products. An efficiency cost function can reflect the number of products in the recommendation results. Therefore, in step 212, an efficiency cost function needs to be set based on the number of recommendations, ensuring that each recommended product incurs a certain "cost," thereby minimizing the inefficiency of an infinite number of recommended products.

[0084] In some embodiments, the efficiency cost function Constructed as: ， in, Let be the efficiency cost vector of type q, where each component represents the efficiency cost of type q for each product to be recommended, i.e., the efficiency cost incurred in recommending that product. Efficiency costs can be defined using different parameters depending on the usage scenario. For example, to help users save money, efficiency costs can be set based on product price; to encourage users to purchase products in bundled packages, efficiency costs can be set based on both bundled packages (including sub-items within bundled packages) and individual items. For instance, selecting three products individually requires consuming three servings, while selecting them as a bundled package only requires consuming one serving, meaning the efficiency cost of individual items is greater than the efficiency cost of a bundled package; efficiency costs can also be set based on factors such as portion size and calories. This application does not limit the specific method for setting efficiency costs.

[0085] By determining the efficiency cost of the recommendation, the number of products in the recommendations can be reasonably limited. In practice, the smaller the efficiency cost function, the more reasonable the recommendation is considered.

[0086] To ensure the product variety in the recommendation results is as diverse as possible and to reduce duplicate recommendations, a group quantity function can be set in step 213. The group quantity function can reflect the product variety in the recommendation results, and the group division method can be based on one or more of the following: product category, price range, and user's historical purchase preferences.

[0087] According to some embodiments of this application, the group quantity function is constructed as follows: ， in, For the number of groups, .

[0088] The groups of products to be recommended can be pre-categorized based on the range of goods. In one example, similar products can be grouped into the same group. This application does not limit the method of grouping products.

[0089] By defining the range of products in the recommendation results, the diversity of the recommendation outcomes can be increased. In practical applications, the larger the number of groups, the higher the degree of diversity in the recommendation results.

[0090] According to some embodiments of this application, reference is made to Figure 3 Step 220 includes step 221: generating a recommendation objective function based on the recommendation scoring function, intent error function, efficiency cost function, and group number function.

[0091] The evaluation of recommendation results can be measured from the perspectives of the recommendation model described above, including scoring, error relative to user needs, efficiency cost, and category diversity. Therefore, a recommendation objective function can be generated based on the recommendation scoring function, intent error function, efficiency cost function, and group quantity function.

[0092] In one example, the recommendation objective function can be generated by maximizing the recommendation rating function, minimizing the intent error function, minimizing the efficiency cost function, and maximizing the number of groups function.

[0093] According to some embodiments of this application, reference is made to Figure 4 Step 221 includes steps 410 to 420.

[0094] Step 410: Based on the predetermined weights of multiple objective functions, the recommendation objective function, intent error function, efficiency cost function, and group number function are weighted respectively to obtain the weighted recommendation objective function, weighted intent error function, weighted efficiency cost function, and weighted group number function.

[0095] Step 420: Generate the recommendation objective function based on the weighted recommendation objective function, the weighted intention error function, the weighted efficiency cost function, and the weighted group quantity function.

[0096] The weights for the objective functions used in the recommendation objective function, intent error function, efficiency cost function, and group quantity function can be set according to different use cases to reflect different value orientations. For example, for use cases that emphasize personal customization, a set of weights can be set for the recommendation objective function, intent error function, efficiency cost function, and group quantity function respectively; for use cases that emphasize optimal price, another set of weights can be set; in addition, for other use cases that emphasize prioritizing key promotional activities, prioritizing diversity, etc., corresponding sets of objective function weights can also be set. This application does not limit the way the objective function weights are set.

[0097] In step 410, a pre-defined set of objective function weights can be used to weight the recommendation objective function, intent error function, efficiency cost function, and group number function, respectively. The weighted recommendation objective function can be expressed as follows: The weighted intention error function can be expressed as: The weighted efficiency cost function can be expressed as: ， The weighted group count function can be expressed as: .in, That is, the objective function weights set for the recommendation objective function, intention error function, efficiency cost function, and group number function.

[0098] According to some embodiments of this application, reference is made to Figure 5 Step 420 includes steps 510 to 520.

[0099] Step 510: Summate the weighted intention error function and the weighted efficiency cost function, and subtract the weighted recommendation objective function and the weighted group number function to obtain the sum recommendation function.

[0100] Step 520: Determine the objective function as minimizing the total recommendation function.

[0101] After obtaining the weighted recommendation objective function, weighted intent error function, weighted efficiency cost function, and weighted group number function, considering maximizing the recommendation rating function, minimizing the intent error function, minimizing the efficiency cost function, and maximizing the group number function, a summative recommendation function can be obtained according to step 510, and minimizing this summative recommendation function can be used as the recommendation objective function. The recommendation objective function obtained in step 520 can be expressed as: 。

[0102] As mentioned above, different weights can be set for different use cases. In some embodiments, these different weights are used to obtain the corresponding recommendation objective function. After solving the function, the recommendation results for different scenarios can be determined, thereby providing users with a variety of options.

[0103] According to some embodiments of this application, reference is made to Figure 6 Step 150 includes steps 610 to 620.

[0104] Step 610: Determine the number of recommendations for each product to be recommended in the decision variables based on the recommendation objective function and recommendation constraints.

[0105] Step 620: Determine the target product from the multiple products to be recommended based on the number of recommendations for each product to be recommended.

[0106] The process of solving for the decision variables based on the recommendation objective function and recommendation constraints is essentially the process of determining each decision component within the decision variables. By determining each decision component, the recommendation quantity for each type of product can be obtained. In the example, the products with a non-zero recommendation quantity are the target products and can be recommended to the user.

[0107] According to some embodiments of this application, step 150 includes: Input the recommended objective function, recommended constraints, and decision variables into the solver; The decision variables are solved using a solver.

[0108] In step 150, a solver can be used to solve the problem. The solver can calculate the solution based on the input data using its built-in algorithm logic, and finally output the solution to the problem. In the embodiments of this application, the recommendation objective function, recommendation constraints, and decision variables can be used as inputs to the solver. The solver then outputs the solution result through its built-in algorithm logic, which is the number of recommendations for each product to be recommended in the decision variables.

[0109] It should be understood that, in addition to using solvers, other methods can be used to solve for decision variables based on the recommended objective function and recommended constraints, such as using large models.

[0110] According to some embodiments of this application, the multiple products to be recommended include at least one multi-level product to be recommended. Each multi-level product to be recommended includes multiple sub-items to be recommended. In this application, "multi-level" refers to two or more levels. The decision component for each multi-level product to be recommended includes a first-level decision component and multiple second-level decision components.

[0111] The first-level decision component is used to indicate the type of product to be recommended at multiple levels.

[0112] Multiple second-level decision components correspond one-to-one with multiple sub-items to be recommended in the multi-level recommended products.

[0113] In some embodiments, a recommended product may contain many optional sub-items. For example, as mentioned above, a recommended product may include individual items and packages. Package-type recommended products may have many optional sub-items, such as a three-piece set with five optional sub-items, from which three need to be selected for ordering. For such multi-level recommended products, in addition to deciding whether to order the package, it is also necessary to decide which specific sub-items to select. Therefore, for multi-level recommended products, a two-level decision component can be set. The first-level decision component represents the package itself, which can also be understood as representing the "package header." In this embodiment, the "package header" indicates the type of package. In different embodiments, the package header can take various forms such as images, text, and numbers, aiming to intuitively present the core information of the package. The second-level decision component represents each optional sub-item within the package. In this embodiment, the "sub-item" indicates the different selectable items within the package, and can also take various forms such as images, text, and numbers, aiming to intuitively present the core information of each selectable item within the package. For example, as mentioned above, for a certain package (ID 1), there are 5 sub-items to choose from (assuming each sub-item corresponds to IDs 2, 3, 4, 5, and 6 respectively). Then, the first-level decision component for this package is... The second-level decision components for each sub-item include .

[0114] When solving for decision variables, if the product to be recommended is a multi-level product in the form of a package, then the first-level decision component and the second-level decision component of the multi-level product to be recommended need to be solved based on the recommendation objective function and constraints, so as to determine whether to recommend the package and which sub-items in the package to recommend.

[0115] According to some embodiments of this application, for any multi-level product to be recommended, step 150 includes the following steps: Based on the recommendation objective function and recommendation constraints, the first-level decision components and multiple second-level decision components of the multi-level recommended products are solved; Based on the first-level decision components obtained from the solution, determine whether the multi-level recommended products are target products; Based on the multiple second-level decision components obtained from the solution, at least one target sub-item is determined from multiple sub-items to be recommended; In response to the multi-level recommended products being the target products, the multi-level recommended products are hierarchically combined based on the first-level decision components and multiple second-level decision components obtained from the solution, so as to obtain a target product including at least one target sub-item.

[0116] Continuing with the example of the package mentioned above (ID 1), the package has 5 sub-items (assuming each sub-item corresponds to IDs 2, 3, 4, 5, and 6 respectively). That is, the decision components for this package include the first-level decision components. Second-level decision components This set meal is a three-piece set, meaning that customers need to choose three items from five sub-items when ordering. Based on the recommendation objective function and constraints mentioned above, these decision components will be solved. If the first-level decision components are obtained... A value of 1 indicates that the package has been identified as the target product and will be recommended to the user. Additionally, if the second-level decision components are obtained... Given the values ​​1, 0, 1, 1, and 0, we can select the target sub-items from these five sub-items, namely the three sub-items with IDs 2, 4, and 5. After determining these three target sub-items, they will be combined based on the first-level and second-level decision components obtained from the solution, forming a package containing the three sub-items with IDs 2, 4, and 5, which will be recommended to the user as the final target product.

[0117] Using this decision-making approach, it is possible to recommend package deals and sub-items within packages, restore the parent-child structure of multi-level products to be recommended, and generate a complete shopping plan.

[0118] After determining the target products based on the recommendation quantity of each product to be recommended, recommendations can be made to the user. In some embodiments, the target products are all individual items, which can be directly recommended to the user. In other embodiments, package deals may also be involved. As mentioned above, the decision variables will include all purchasable products in the menu, such as all individual items and all package deals (and all possible sub-items included in the package deals). That is, in the decision variables, for the case where the package deal includes sub-items, the package deal header + sub-item structure can be used as the decision component in the decision variables for solution, that is, the first-level decision component and the second-level decision component. After obtaining the solution result, if the target product involves package-related products, the parent-child structure in the package deal can be restored according to the package deal header + sub-item structure, thereby obtaining the final recommendation scheme output to the customer, realizing cross-granularity recommendation, enabling the recommendation method to support the combination recommendation of individual items, packages and sub-items, effectively improving the flexibility of the recommendation.

[0119] Figure 10 An embodiment of an agent-based recommendation method is illustrated herein. It should be understood that this embodiment is described for illustrative purposes only and should not be construed as limiting the scope of this application.

[0120] like Figure 10 As shown, in step 1010, the user's demand intent can be obtained, which can be represented in the form of intent parameters. In step 1020, the demand intent in the form of intent parameters can be rewritten, for example, rewriting the user's demand quantity as the product demand quantity, correcting and updating the demand intent, and determining the parameter range corresponding to the demand intent. In step 1030, decision variables can be generated. .

[0121] In step 1040, recommendation constraints can be generated based on the relevant data of the product to be recommended. The recommendation constraints can be any of the following: ， ， 。

[0122] In step 1050, a recommendation objective function can be generated based on the rating results of the recommendation model and relevant data of the product to be recommended. The recommendation objective function can be expressed as follows: 。

[0123] In step 1060, the recommendation objective function, recommendation constraints, and decision variables can be used as inputs to the solver, which solves the problem to obtain the recommendation quantity of each product to be recommended, thereby determining the target product to be recommended to the user from the products to be recommended.

[0124] For recommendations involving packages, in step 1070, a solution can be assembled. The parent-child structure in the package is restored based on the structure of package header + sub-items, so as to obtain the final recommended solution output to the customer.

[0125] Based on the same technical concept, embodiments of this application provide an agent-based recommendation device. Embodiments of the agent-based recommendation device can be referenced from embodiments of the agent-based recommendation method; repeated details will not be repeated. Reference Figure 8 The agent-based recommendation device 800 includes an acquisition module 810, a first generation module 820, a second generation module 830, a third generation module 840, a solution module 850, and a recommendation module 860.

[0126] The acquisition module 810 is used to acquire at least one requirement intent from user input.

[0127] The first generation module 820 is used to generate decision variables based on multiple products to be recommended. The decision variables include decision components for each product to be recommended.

[0128] The second generation module 830 is used to generate recommendation constraints based on the recommendation conditions of multiple products to be recommended.

[0129] The third generation module 840 is used to generate a recommendation objective function based on at least one demand intent.

[0130] The solver module 850 is used to solve the decision variables based on the recommendation objective function and recommendation constraints in order to determine the target product from multiple products to be recommended.

[0131] The recommendation module 860 is used to recommend target products to users.

[0132] The acquisition module 810, first generation module 820, second generation module 830, third generation module 840, solution module 850, and recommendation module 860 in the agent-based recommendation device 800 can correspond to steps 110 to 160 in the agent-based recommendation method 100. For the sake of brevity, they will not be described in detail here. It should be understood that, corresponding to the embodiment of the agent-based recommendation method 100, the embodiment of the agent-based recommendation device 800 may also include more modules.

[0133] It should be noted that the functions of the modules discussed herein can be divided into multiple modules, and / or at least some functions of multiple modules can be combined into a single module. The specific actions performed by a particular module discussed herein include the specific module itself performing the action, or alternatively, the specific module calling or otherwise accessing another component or module that performs the action (or performs the action in conjunction with the specific module). Therefore, a specific module performing an action can include the specific module performing the action itself and / or another module that performs the action, called or otherwise accessed by the specific module.

[0134] It should also be understood that this article can describe various technologies in the general context of software and hardware components or program modules. The above regarding... Figure 8 The described modules can be implemented in hardware or in hardware in combination with software and / or firmware. For example, these modules can be implemented as computer program code / instructions configured to execute in one or more processors and stored in a computer-readable storage medium. Alternatively, these modules can be implemented as hardware logic / circuit. Hardware logic / circuit may include integrated circuit chips (which include processors (e.g., Central Processing Unit (CPU), microcontrollers, microprocessors, digital signal processors (DSPs), etc.), memory, one or more communication interfaces, and / or one or more components in other circuitry), and may optionally execute received program code and / or include embedded firmware to perform functions.

[0135] This application provides a computing device 900, such as... Figure 9 As shown. Figure 9 An example configuration of a computing device 900 that can be used to implement the agent-based recommendation method 100 described herein is shown. For example, the agent-based recommendation apparatus 800 described above can be implemented wholly or at least partially by the computing device 900 or a similar device or system.

[0136] The computing device 900 may include at least one processor 905 capable of communicating with each other, such as via a bus 904 or other suitable connection, a memory 907, multiple communication interfaces 902, a display device 901, other input / output (I / O) devices 903, and one or more mass storage devices 906. Instructions are stored on the memory 907 that, when executed by the processor 905, cause the processor 905 to perform the agent-based recommended method as described in the above embodiments.

[0137] The computing device 900 can be a variety of different types of devices. Examples of the computing device 900 include, but are not limited to: desktop computers, server computers, laptop or netbook computers, mobile devices (e.g., tablets, cellular or other wireless phones (e.g., smartphones), notebook computers, mobile stations), wearable devices (e.g., glasses, watches), entertainment devices (e.g., entertainment appliances, set-top boxes communicatively coupled to a display device, game consoles), televisions or other display devices, automotive computers, and so on.

[0138] Processor 905 may be a single processing unit or multiple processing units, and all processing units may include single or multiple computing units or multiple cores. Processor 905 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuits, and / or any device that manipulates signals based on operating instructions. Among other capabilities, processor 905 may be configured to fetch and execute computer-readable instructions stored in memory 907, mass storage device 906, or other computer-readable media, such as program code of operating system 908, program code of application program 909, program code of other program 910, etc.

[0139] Memory 907 and mass storage device 906 are examples of computer-readable storage media for storing instructions that are executed by processor 905 to perform the various functions described above. For example, memory 907 can generally include both volatile and non-volatile memory (e.g., RAM, ROM, etc.). Furthermore, mass storage device 906 can generally include hard disk drives, solid-state drives, removable media, including external and removable drives, memory cards, flash memory, floppy disks, optical disks (e.g., CDs, DVDs), storage arrays, network-attached storage, storage area networks, etc. Both memory 907 and mass storage device 906 can be collectively referred to herein as memory or computer-readable storage media, and can be non-transitory media capable of storing computer-readable, processor-executable program instructions as computer program code, which can be executed by processor 905 as a specific machine configured to perform the operations and functions described in the examples herein.

[0140] Multiple programs can be stored on mass storage device 906. These programs include operating system 908, one or more application programs 909, other programs 910, and program data 911, and they can be loaded into memory 907 for execution. Examples of such application programs or program modules may include, for example, computer program logic (e.g., computer program code or instructions) for implementing the following components / functions: agent-based recommendation device 800 (including acquisition module 810, first generation module 820, second generation module 830, third generation module 840, solution module 850, and recommendation module 860), agent-based recommendation method 100 (including any suitable steps of agent-based recommendation method 100), and / or other embodiments described herein.

[0141] Although Figure 9 The data is illustrated as being stored in memory 907 of computing device 900, but operating system 908, application program 909, other programs 910 and program data 911 or portions thereof may be implemented using any form of computer-readable medium accessible by computing device 900.

[0142] One or more communication interfaces 902 are used for exchanging data with other devices, such as via a network, direct connection, etc. Such communication interfaces can be one or more of the following: any type of network interface (e.g., a network interface card (NIC)), wired or wireless (such as IEEE 802.11 Wireless LAN (WLAN)) wireless interface, Wi-MAX interface, Ethernet interface, Universal Serial Bus (USB) interface, cellular network interface, Bluetooth™ interface, Near Field Communication (NFC) interface, etc. Communication interface 902 can facilitate communication across various network and protocol types, including wired networks (e.g., LAN, cable, etc.) and wireless networks (e.g., WLAN, cellular, satellite, etc.), the Internet, etc. Communication interface 902 can also provide communication with external storage devices (not shown), such as storage arrays, network-attached storage, storage area networks, etc.

[0143] In some examples, a display device 901, such as a monitor, may be included for displaying information and images to the user. Other I / O devices 903 may be devices that receive various inputs from the user and provide various outputs to the user, and may include touch input devices, gesture input devices, cameras, keyboards, remote controls, mice, printers, audio input / output devices, and so on.

[0144] The technologies described herein can be supported by these various configurations of computing device 900, and are not limited to specific examples of the technologies described herein. For example, the functionality can also be implemented wholly or partially on a “cloud” using a distributed system. A cloud includes and / or represents a platform for resources. The platform abstracts the underlying functionality of the cloud’s hardware (e.g., servers) and software resources. Resources may include applications and / or data that can be used when performing computational processing on servers remote from computing device 900. Resources may also include services provided via the Internet and / or via subscriber networks such as cellular or Wi-Fi networks. The platform can abstract resources and functionality to connect computing device 900 to other computing devices. Therefore, the implementation of the functionality described herein can be distributed throughout the cloud. For example, the functionality can be implemented partly on computing device 900 and partly through a platform that abstracts the functionality of the cloud.

[0145] This application also provides a computer-readable storage medium storing instructions that, when executed individually or jointly by one or more processors of a computing device, cause the computing device to perform the methods described in any of the above embodiments.

[0146] Computer-readable storage media include volatile and non-volatile, removable and non-removable media implemented by any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, Digital Universal Disc (DVD) or other optical storage devices, magnetic cassettes, magnetic tapes, disk storage devices or other magnetic storage devices, or any other non-transmission medium that can be used to store information for access by computer equipment.

[0147] This application also provides a computer program product including instructions that, when executed individually or jointly by one or more processors of a computing device, cause the computing device to perform the methods as described in any of the above embodiments.

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and not to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and they should all be covered within the scope of the claims and specification of this application. In particular, as long as there is no structural conflict, the various technical features mentioned in the embodiments can be combined in any way. This application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. An agent-based recommendation method, comprising: Obtain at least one user intent from their input; Based on multiple products to be recommended, decision variables are generated, and the decision variables include decision components for each product to be recommended. Based on the recommendation criteria of the multiple products to be recommended, generate recommendation constraints; Based on the at least one stated demand intent, a recommendation objective function is generated; The decision variables are solved based on the recommendation objective function and the recommendation constraints to determine the target product from the plurality of products to be recommended. as well as The target product is recommended to the user.

2. The recommendation method according to claim 1, wherein, The generation of the recommendation objective function based on the at least one demand intent includes: At least one recommendation model is used to determine the recommendation score corresponding to the decision variable, so as to generate a recommendation score function corresponding to each of the at least one recommendation model; and The recommendation objective function is generated based at least on the recommendation scoring function.

3. The recommendation method according to claim 2, wherein, The generation of the recommendation objective function based on the at least one demand intent further includes: Based on the at least one demand intent, generate the intent error function corresponding to the decision variable; Based on the number of recommendations for each product to be recommended, generate the efficiency-cost function corresponding to the decision variables; and Based on the product range of each product to be recommended, generate a function for the number of groups corresponding to the decision variables.

4. The recommended method according to claim 3, wherein, Generating the recommendation objective function based at least on the recommendation scoring function includes: The recommendation objective function is generated based on the recommendation scoring function, the intent error function, the efficiency cost function, and the group number function.

5. The recommended method according to claim 4, wherein, The step of generating the recommendation objective function based on the recommendation scoring function, the intent error function, the efficiency cost function, and the group number function includes: Based on predetermined weights for multiple objective functions, the recommendation objective function, the intent error function, the efficiency cost function, and the group quantity function are weighted respectively to obtain a weighted recommendation objective function, a weighted intent error function, a weighted efficiency cost function, and a weighted group quantity function; and The recommendation objective function is generated based on the weighted recommendation objective function, the weighted intention error function, the weighted efficiency cost function, and the weighted group number function.

6. The recommended method according to claim 5, wherein, The step of generating the recommendation objective function based on the weighted recommendation objective function, the weighted intention error function, the weighted efficiency cost function, and the weighted group number function includes: The weighted intention error function and the weighted efficiency cost function are summed, and the weighted recommendation objective function and the weighted group quantity function are subtracted to obtain the summed recommendation function; and The objective function for recommendation is determined to be minimizing the sum recommendation function.

7. The recommended method according to claim 3, wherein, The plurality of products to be recommended includes n products, where n is a natural number greater than 1, and the decision variables are constructed as follows: ， in, As decision variables, Each of the n products to be recommended corresponds to one of them and indicates the number of recommended products for each product.

8. The recommended method according to claim 7, wherein, The recommended constraints are constructed as one of the following: ， ， ， in, Let j be the coefficient vector corresponding to the j-th constraint. Let be the constraint value that the j-th constraint needs to satisfy.

9. The recommended method according to claim 7, wherein, The intent error function is constructed as one of the following: ， ， ， in, Let i be the intent error function corresponding to the i-th demand intent. It is an absolute value function. , Let i be the coefficient vector corresponding to the i-th demand intention. This represents the demand value corresponding to the i-th demand intent.

10. The recommendation method according to claim 9, wherein, The coefficient vector corresponding to the i-th demand intention The components in can be constructed as: ， in, For the p-th decision component in the decision variables The corresponding coefficients, Represents the p-th decision component Does it satisfy the i-th requirement? Represents the p-th decision component The measure of the i-th demand intention.

11. The recommendation method according to claim 7, wherein, The group quantity function is constructed as follows: ， in, For the number of groups, .

12. The recommendation method according to claim 7, wherein, The step of solving for the decision variables based on the recommendation objective function and the recommendation constraints to determine the target product from the plurality of products to be recommended includes: Based on the recommendation objective function and the recommendation constraints, determine the recommendation quantity for each product to be recommended among the decision variables; and The target product is determined from the plurality of products to be recommended based on the number of recommendations for each product to be recommended.

13. The recommended method according to any one of claims 1-12, wherein, The step of solving for the decision variables based on the recommendation objective function and the recommendation constraints to determine the target product from the plurality of products to be recommended includes: The recommended objective function, the recommended constraints, and the decision variables are input into the solver; and The decision variables are solved using the solver.

14. The recommended method according to any one of claims 1-12, wherein, The at least one demand intent includes at least one of the following: user demand quantity, recommended product range demand, product demand quantity, product price demand, and user behavior preference.

15. The recommended method according to claim 14, further comprising at least one of the following: The quantity of products requested by the user is determined based on the quantity of user demand and the user's historical intent data; The range of recommended products input by the user is updated based on the product parameters of the multiple products to be recommended. Based on the at least one requirement intent, determine the parameter range corresponding to any one of the at least one requirement intents.

16. The recommendation method according to claim 1, wherein, The plurality of products to be recommended includes at least one multi-level product to be recommended, and each multi-level product to be recommended includes a plurality of sub-items to be recommended. The decision components for each multi-level product to be recommended include: The first-level decision component is used to indicate the type of the multi-level recommended product; and Each of the multiple second-level decision components corresponds one-to-one with the multiple sub-items to be recommended included in the multi-level recommended products.

17. The recommendation method according to claim 16, wherein, The step of solving for the decision variables based on the recommendation objective function and the recommendation constraints to determine the target product from the plurality of products to be recommended includes: For any of the aforementioned multi-level recommended products: Based on the recommendation objective function and the recommendation constraints, the first-level decision components and the plurality of second-level decision components of the multi-level recommended products are solved; Based on the first-level decision components obtained from the solution, determine whether the multi-level products to be recommended are the target products; Based on the multiple second-level decision components obtained from the solution, at least one target sub-item is determined from the multiple sub-items to be recommended; and In response to the fact that the multi-level recommended products are the target products, the multi-level recommended products are hierarchically combined based on the solved first-level decision components and the solved multiple second-level decision components to obtain the target product including the at least one target sub-item.

18. A recommendation device based on an intelligent agent, comprising: The acquisition module is used to acquire at least one user intent input. The first generation module is used to generate decision variables based on multiple products to be recommended, wherein the decision variables include decision components for each product to be recommended; The second generation module is used to generate recommendation constraints based on the recommendation conditions of the multiple products to be recommended; The third generation module is used to generate a recommendation objective function based on the at least one demand intent; The solution module is used to solve the decision variables according to the recommendation objective function and the recommendation constraints, so as to determine the target product from the plurality of products to be recommended; as well as The recommendation module is used to recommend the target product to the user.

19. A computing device, comprising: At least one processor; as well as At least one memory communicatively connected to the at least one processor, the at least one memory storing instructions that, when executed individually or jointly by the at least one processor, cause the computing device to perform the recommended method of any one of claims 1 to 17.

20. A computer-readable storage medium storing instructions that, when executed individually or jointly by one or more processors of a computing device, cause the computing device to perform the recommended method of any one of claims 1 to 17.

21. A computer program product comprising instructions that, when executed individually or jointly by one or more processors of a computing device, cause the computing device to perform the recommended method of any one of claims 1 to 17.

Images