Category optimization method, system and storage medium based on shopping basket purchase behavior

Abstract

This invention discloses a category optimization method, system, and storage medium based on shopping basket purchase behavior. The category optimization method steps are as follows: S1: Obtain historical transaction data and estimate consumer preferences; S2: Group products according to revenue; S3: Extract historical data and optimize the categories of each product group using an optimization model; S4: Select the optimized product group with the highest expected revenue as the final category recommendation; S5: Send the final category recommendation results to the management system for intelligent category display decision-making. This invention not only improves the computational efficiency of category optimization, but also effectively controls the operational burden of enterprises, ensuring service quality and customer satisfaction.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, system, and storage medium for optimizing product categories based on shopping basket purchase behavior. Background Art

[0002] With the advent of the big data era, in-depth analysis of consumer shopping basket purchase behavior has become a crucial issue in enterprise category optimization. Traditional retailers' category management (or product selection management) primarily relies on the experience and intuition of sales staff, or adopts a long-tail strategy, offering consumers all available products. In practice, this approach, due to limited human resources, often overlooks product dependencies and consumer purchasing habits, resulting in an irrational product mix and room for profit improvement. Furthermore, incomplete information or human misjudgment can lead to relatively accurate demand forecasts for some products, while forecasts for others exhibit significant deviations, resulting in fluctuations in overall accuracy. Therefore, improving the utilization of shopping basket data, enhancing the accuracy of product demand forecasts, and, consequently, optimizing the efficiency and accuracy of category management, is a pressing issue.

[0003] The patent with publication number CN118735634A discloses a method, device and computing equipment for processing product listings. It specifically discloses a method for building a product resource pool by collecting massive amounts of product data, and screening candidate products from the product selection pool based on the user's key product selection information, so that the user can select products from the candidate products. The patent with publication number CN118735638A discloses a product selection method and system for a cross-border e-commerce platform based on big data: generating a market trend set; obtaining and recording the sales of merchants' historical product selections; analyzing market trends and evaluating merchants' historical product selections; building a product selection analysis model, analyzing the candidate products provided by merchants based on the model, and generating a product selection recommendation strategy. Although the above patent has achieved a certain intelligent product selection decision-making method based on historical data, it does not take into account consumers' shopping basket purchase behavior.

[0004] In addition, existing online convex optimization methods are widely used in the field of machine learning, but are rarely used in the field of integer optimization, such as category optimization. Patent publication number CN117668712A discloses a category optimization method and related device based on artificial intelligence and the Nine-Palace Screening Method: historical operating data is obtained and model training is performed to obtain a target category optimization model set; the target category optimization model set is used to perform comprehensive indicator calculations on the multi-dimensional indicators and multiple indicators of the historical operating data to obtain an indicator data set; the indicator data set is optimized and calculated using the Nine-Palace Screening Algorithm, and a product list for each grid is output. Category optimization recommendation data is generated based on the product list for each grid; and the category optimization recommendation data is optimized and interactively output with report output based on the screening conditions to obtain multiple target product list reports. The patent with publication number CN115204506A discloses a method, device and computer equipment for optimizing the category of power equipment, which determines the category to be evaluated and multiple evaluation items; calculates the evaluation parameters corresponding to each evaluation item based on the statistical data of the category to be evaluated; determines the optimization parameters of the category to be evaluated based on each evaluation parameter; and determines that the category to be evaluated needs to be optimized when the optimization parameters meet the optimization conditions. This patented method can reduce the influence of subjective judgment factors in the optimization of production equipment categories and achieve objective quantitative evaluation of equipment category optimization work. However, the above-mentioned category optimization method does not have the characteristics of an e-commerce environment and does not have convergent results; the method does not take the retailer's maximum profit as the optimization objective function, and cannot solve the category optimization problem faced by e-commerce companies that expect to maximize profits. Summary of the Invention

[0005] Based on the technical problems existing in the background technology, the present invention proposes a category optimization method, system and storage medium based on shopping basket purchase behavior, which not only improves the computational efficiency of category optimization, but also effectively controls the operating burden of the enterprise, ensuring the service level and customer satisfaction of the enterprise.

[0006] The category optimization method based on shopping basket purchase behavior proposed in the present invention has the following steps:

[0007] S1: Obtain historical transaction data and estimate consumer preferences;

[0008] S2: Group products by revenue;

[0009] S3: Extract historical data and optimize the product categories of each group through the optimization model;

[0010] S4: Select the product group optimization result with the highest expected profit as the final category recommendation;

[0011] S5: Send the final category recommendation results to the management system for intelligent category display decision-making.

[0012] Preferably, the consumer preferences in S1 include consumer type and quantity purchased by product With preference list Composition and each stage The arrival probability corresponding to each consumer type .

[0013] Preferably, the grouping method in S2 is: sort the product set that has been pre-selected and optimized according to net profit, and group them according to a certain multiple, and the maximum profit difference in each group of products does not exceed the multiple.

[0014] Preferably, the optimization model in S3 is:

[0015]

[0016] in, Purchase quantity for the product; is a preference list; is the arrival probability; Data for each stage; A collection of data from all historical stages; For products; For consumer type; is the total number of all consumer types; is a binary decision variable, i.e. whether to provide the product.

[0017] Preferably, the method steps for category optimization in S3 are as follows:

[0018] S31: Set the starting solution of the optimization model to , on each product group, loop t=1,2,…,T;

[0019] S32: Extract historical transaction data using independent and identical distribution without replacement;

[0020] S33: Calculate the super gradient through the optimization model based on the extracted historical transaction data;

[0021]

[0022] S34: Update ;

[0023] S35: Euclidean projection into the feasible region, ;

[0024] S36: By rounding to the interval {0,1}, we get ;

[0025] S37: For all Take the average and then use the rounding method to get an integer solution as the category optimization result.

[0026] Preferably, the rounding method is a random exchange rounding method, a random pipeline rounding method or a dependent rounding method.

[0027] The category optimization system based on shopping basket purchase behavior proposed in the present invention includes:

[0028] Acquisition module, used to obtain historical transaction data;

[0029] Estimation module, used to estimate consumer preferences;

[0030] The grouping module is used to group products according to their revenue;

[0031] The optimization module is used to extract historical data and optimize the category of each group of products through the optimization model;

[0032] The push module is used to select the product group optimization results with the highest expected profit as the final category recommendation, and send the final category recommendation results to the management system for intelligent category display decision-making.

[0033] The computer-readable storage medium proposed in the present invention stores a computer program, and when the computer program is executed by a processor, the above-mentioned category optimization method based on shopping basket purchase behavior is implemented.

[0034] Beneficial technical effects of the present invention:

[0035] Unlike the existing technology that uses human experience to select products and artificial intelligence to predict, the present invention adopts a more fine-grained customer demand modeling method, further considering consumers' multi-product purchasing behavior and preference modeling methods that conform to rational consumer choice behavior. Based on maximizing the retailer's profits, the product category optimization results are obtained, which improves consumer satisfaction while reducing the retailer's operating costs, ensuring that the product types owned are within a reasonable range and ensuring the company's operating efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flow chart of the category optimization method based on shopping basket purchase behavior proposed by the present invention. DETAILED DESCRIPTION

[0037] The present invention will be further explained below with reference to specific embodiments.

[0038] Example 1

[0039] Reference Figure 1 , the present invention proposes

[0040] The category optimization method based on shopping basket purchase behavior proposed in the present invention has the following steps:

[0041] S1: Obtain historical transaction data and estimate consumer preferences

[0042] Set conditional information and filter out historical sales data that meet the conditions, where the conditional information includes the historical time range and the complete set of products that need category optimization. Based on the above historical data, the estimated consumer preferences include: Consumer type and quantity purchased by product With preference list Composition and each stage The arrival probability corresponding to each consumer type .

[0043] S2: Group products by revenue

[0044] Specifically: the product collection that will be optimized in advance According to net profit Sort from largest to smallest, Indicates product net profit; group them according to multiples of 2, making the total number of groups equal to M, and ensure that the maximum profit difference in each group of products does not exceed 2.

[0045] S3: Extract historical data and optimize the product categories for each group through the optimization model

[0046] Among them, the optimization model is:

[0047]

[0048] in, Purchase quantity for the product; is a preference list; is the arrival probability; Data for each period (day, week, etc.); A collection of data from all historical stages; For products; For consumer type; is the total number of all consumer types; It is a binary decision variable, i.e. whether to provide the product. It takes 1 when the product is provided and 0 when the product is not provided.

[0049] The optimization steps are as follows:

[0050] S31: Input the estimated results based on historical data; product grouping results; learning rate of each group m ;

[0051] S32: Set the starting solution to , on each product group, loop t=1,2,…,T;

[0052] S33 extracts historical transaction data using independent and identical distribution without replacement;

[0053] S34: Calculate super gradient based on the extracted historical transaction data ;

[0054] S35: Update ;

[0055] S36: Euclidean projection into the feasible region, ;

[0056] S37: By rounding to the interval {0,1}, we can get ;

[0057] S37: For all Take the average and then use the rounding method to get an integer solution as the category optimization result.

[0058] The rounding method used is random exchange rounding, random pipeline rounding or dependent rounding.

[0059] S4: Select the product group with the highest expected profit as the final category recommendation

[0060] Specifically, initialize the total profit to 0; for each customer type k, perform the following steps:

[0061] Find out which products are available (those that meet certain criteria);

[0062] Select the first few available products (quantity );

[0063] Calculate the profits of these selected products and add them to the total profit;

[0064] Returns the calculated total profit.

[0065] S5: Send the final category recommendation results to the management system for intelligent category display decision-making.

[0066] This invention has demonstrated excellent performance in the retailer decision-making process. For example, in e-commerce companies, the system provides sales personnel with a responsive and scientific category optimization tool, improving the level of category optimization. For businesses, this can increase order volume and transaction amounts, reduce personnel costs, and maintain customer satisfaction. Specifically, in simulations, this method achieved a near-optimal solution for 100 products in less than 2 seconds, 97% of the results achieved by a commercial solver in 1800 seconds.

[0067] The present invention adopts a more fine-grained customer demand modeling method, further considers consumers' multi-product purchasing behavior and preference modeling methods that conform to rational consumer choice behavior, and obtains product category optimization results based on maximizing retailers' profits, thereby improving consumer satisfaction while reducing retailers' operating costs, ensuring that the product variety owned is within a reasonable range, and ensuring the company's operating efficiency; the results of the embodiment show that the method of the present invention has good computing performance and practical application value.

[0068] Example 2

[0069] The category optimization system based on shopping basket purchase behavior proposed in the present invention includes:

[0070] Acquisition module, used to obtain historical transaction data;

[0071] Estimation module, used to estimate consumer preferences;

[0072] The grouping module is used to group products according to their revenue;

[0073] The optimization module is used to extract historical data and optimize the category of each group of products through the optimization model;

[0074] The push module is used to select the product group optimization results with the highest expected profit as the final category recommendation, and send the final category recommendation results to the management system for intelligent category display decision-making.

[0075] Example 3

[0076] The computer-readable storage medium proposed in the present invention stores a computer program, and when the computer program is executed by a processor, the category optimization method based on shopping basket purchase behavior of Example 1 is implemented.

[0077] The category optimization method of the present invention can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Various aspects of the present invention can be implemented as machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, RAM, ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the process described herein. In addition, the machine-readable code, or part thereof, can be transmitted over a wired or wireless network. When such media includes instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor, the invention described herein includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques of the present invention, the present invention also includes the computer itself.

Claims

1. Category optimization method based on shopping basket purchase behavior, characterized by: The steps are as follows: S1: Obtain historical transaction data and estimate consumer preferences; S2: Group products by revenue; S3: Extract historical data and optimize the product categories of each group through the optimization model; S4: Select the product group optimization result with the highest expected profit as the final category recommendation; S5: Send the final category recommendation results to the management system for intelligent category display decision-making; The consumer preferences in S1 include the consumer type k and the quantity q purchased by the product. k With the preference list σ k Composed of and corresponding to each consumer type in each stage t arrival probability The optimized model in S3 is: Where T is the set of all historical stages; i is the product; K is the total number of all consumer types; x i is a binary decision variable, i.e., whether to provide the product; The steps for category optimization in S3 are as follows: S31: Input the estimated results based on historical data; product grouping results; learning rate η for each group m m ; Set the starting solution of the optimization model to x1 = y1 = 0, and loop t = 1, 2, ..., T for each product group; S32: Extract historical transaction data using independent and identical distribution without replacement; S33: Calculate the super gradient through the optimization model based on the extracted historical transaction data; S34: Update S35: Euclidean projection into the feasible region, S36: By rounding to the interval {0,1}, we get S37: For all Take the average and then use the rounding method to get an integer solution as the category optimization result.

2. The category optimization method based on shopping basket purchase behavior according to claim 1, characterized in that: The grouping method in S2 is: sort the product set that has been pre-selected and optimized according to net profit, and group them according to a certain multiple, and the maximum profit difference in each group of products does not exceed the multiple.

3. The category optimization method based on shopping basket purchase behavior according to claim 1, characterized in that: The rounding method is random exchange rounding, random pipeline rounding, or dependent rounding.

4. Category optimization system based on shopping basket purchase behavior, characterized by: include: Acquisition module, used to obtain historical transaction data; Estimation module, used to estimate consumer preferences; The grouping module is used to group products according to their revenue; The optimization module is used to extract historical data and optimize the category of each group of products through the optimization model; The push module is used to select the product group optimization results with the highest expected profit as the final category recommendation and send the final category recommendation results to the management system for intelligent category display decision-making; Consumer preferences include consumer type k and product purchase quantity q k With the preference list σ k Composed of and corresponding to each consumer type in each stage t arrival probability The optimization model is: Where T is the set of all historical stages; i is the product; K is the total number of all consumer types; x i is a binary decision variable, i.e., whether to provide the product; The steps for category optimization are as follows: Input the estimated results based on historical data; product grouping results; learning rate η for each group m m ; Set the starting solution of the optimization model to x1 = y1 = 0, and loop t = 1, 2, ..., T for each product group; Extract historical trading data using independent and identically distributed methods without replacement; Based on the extracted historical transaction data, the super gradient is calculated through the optimization model; renew Will Euclidean projection into the feasible region, Will By rounding to the interval {0,1}, we get For all Take the average and then use the rounding method to get an integer solution as the category optimization result.

5. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the category optimization method based on shopping basket purchase behavior as described in any one of claims 1 to 3 is implemented.

Images