Shop grading method and electronic equipment

Abstract

The embodiment of the invention provides a shop grade division method and electronic equipment, and the method comprises the steps: obtaining a shop data set of each shop in a plurality of shops of an e-commerce platform, and constructing a data feature set of each shop according to the shop data set; the data feature set is input into a preset income prediction model, income prediction is conducted on the data feature set through the income prediction model, multiple types of income prediction values are obtained, and the income prediction model comprises multiple income prediction modules corresponding to the multiple types; each income prediction module is used for performing income prediction of a corresponding category on the data feature set to obtain an income prediction value of the corresponding category; and determining a total income prediction value of the corresponding shop based on the income prediction values of all categories, and determining a value level of each shop in the e-commerce platform according to the total income prediction value of all shops. According to the invention, the prediction accuracy of the income of the shop can be improved, and the accuracy of carrying out value grade division on the shop according to the predicted income of the shop is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of artificial intelligence, relates to deep learning technology, and in particular to a store grade classification method and an electronic device. BACKGROUND

[0002] The competition between stores is intensified due to the large number of stores on a cross-border e-commerce platform. The traditional store operation mode (such as uniformly allocating traffic, marketing resources, or formulating risk control strategies) cannot meet the differentiated needs, resulting in problems such as resource waste, inefficient marketing, and lagging risks. It is urgent to realize the fine management of stores by classifying the value of stores.

[0003] However, due to the single store value evaluation method, unreasonable evaluation indexes, and weak dynamic adaptability in the related art, the accuracy of store classification is insufficient, and it is difficult to provide a scientific basis for the platform to allocate resources, formulate marketing strategies, and control risks for stores. SUMMARY

[0004] In view of the above, it is necessary to provide a store grade classification method and an electronic device, which can solve the problem of resource waste caused by the difficulty in accurately classifying stores.

[0005] The first aspect of the embodiment of the application provides a store grade classification method, including: obtaining a store data set of each store in a plurality of stores of an e-commerce platform, and constructing a data feature set of each store according to the store data set; inputting the data feature set into a preset income prediction model, and using the income prediction model to perform income prediction on the data feature set to obtain income prediction values of a plurality of categories, wherein the income prediction model includes a plurality of income prediction modules corresponding to the plurality of categories, each income prediction module is used for performing income prediction of a corresponding category on the data feature set to obtain an income prediction value of the corresponding category; determining an income prediction total value of a corresponding store based on the income prediction values of all categories, and determining a value grade of each store in the e-commerce platform according to the income prediction total values of all stores.

[0006] According to an embodiment of the present application, the store data set includes basic data of a store, subscription income data, commission income data, and payment return commission data, and the method of obtaining the data feature set of each store according to the store data set includes: performing data cleaning processing and format standardization processing on the store data set to obtain a preprocessed store data set; determining a plurality of data features of each store according to the preprocessed store data set, wherein the plurality of data features include basic features corresponding to the basic data, subscription income features corresponding to the subscription income data, commission income features corresponding to the commission income data, and payment return commission features corresponding to the payment return commission data; and constructing the data feature set according to the plurality of data features.

[0007] According to an embodiment of the present application, the plurality of categories include a subscription income category, a commission income category, and a payment return commission category, and the method further includes: collecting a historical store data set of each store, constructing a historical data feature set of each store according to the historical store data set, wherein the historical data feature set includes historical basic features, historical subscription income features, historical commission income features, and historical payment return commission features; constructing a sample data set according to the historical store data set and the corresponding historical data feature set, and dividing the sample data set into a training set and a validation set; and training an income prediction module corresponding to each category in the income prediction model by using data features of each category in the training set.

[0008] According to an embodiment of the present application, the method further includes: training an income prediction module corresponding to a commission income category in the income prediction model by using historical commission income features in the training set, including: constructing a regression model based on a LightGBM algorithm, determining an importance score of each feature in the historical commission income features in a training process of the regression model, and determining core features in the historical commission income features according to the importance score; training the regression model based on the core features, so that the regression model learns a nonlinear mapping relationship between the core features and the income of the commission income category, and obtains the trained income prediction module corresponding to the commission income category.

[0009] According to an embodiment of the present application, the method further includes: determining a pseudo Gini coefficient corresponding to the income prediction module, and determining that the income prediction module is trained if the pseudo Gini coefficient is greater than a preset coefficient threshold.

[0010] According to an embodiment of this application, determining the pseudo-Gini coefficient corresponding to the revenue prediction module includes: sorting the multiple stores in descending order of the revenue prediction values ​​predicted by the revenue prediction module; determining the cumulative actual revenue percentage corresponding to multiple quantile positions in the sorted sequence based on the actual revenue value of each store after sorting, to obtain a first cumulative actual percentage; sorting the multiple stores in a random order, and determining the cumulative actual revenue percentage corresponding to multiple quantile positions in the random sorted sequence based on the actual revenue value of each store after random sorting, to obtain a second cumulative actual percentage; drawing a first proportional curve based on the first cumulative actual percentage, where the horizontal axis of the first proportional curve indicates the quantile and the vertical axis indicates the first cumulative actual percentage; drawing a second proportional curve based on the second cumulative actual percentage, where the horizontal axis of the second proportional curve indicates the quantile and the vertical axis indicates the second cumulative actual percentage; determining the area between the first proportional curve and the second proportional curve, and determining the pseudo-Gini coefficient based on the area.

[0011] According to an embodiment of this application, after determining that the income prediction module has completed training, the method further includes: determining the updated pseudo-Gini coefficient corresponding to the income prediction module based on the validation set; if the updated pseudo-Gini coefficient is less than or equal to a preset coefficient threshold, retraining the income prediction module.

[0012] According to an embodiment of this application, determining the revenue level of each store on the e-commerce platform based on the total predicted revenue of all stores includes: sorting the multiple stores in descending order of the total predicted revenue; determining the cumulative predicted revenue percentage corresponding to multiple quantile positions in the sorted sequence based on the total predicted revenue of each store after sorting; plotting a predicted revenue percentage curve based on the cumulative predicted revenue percentage, where the horizontal axis of the predicted revenue percentage curve indicates the quantile and the vertical axis indicates the cumulative predicted revenue percentage; determining the curvature of the predicted revenue percentage curve and determining the division position based on the quantile corresponding to the maximum curvature; classifying the stores located before the division position among the sorted stores as high-value stores; and classifying the stores other than the high-value stores as medium-value stores and low-value stores based on the pyramid division rule.

[0013] According to an embodiment of this application, after determining the value level of each store on the e-commerce platform, the method further includes: determining corresponding platform resources for the multiple stores based on the value level.

[0014] A second aspect of this application provides a store rating device, comprising: a data processing module for acquiring a store dataset for each store among multiple stores on an e-commerce platform, and constructing a data feature set for each store based on the store dataset; a model prediction module for inputting the data feature set into a preset revenue prediction model, and using the revenue prediction model to predict revenue for the data feature set to obtain revenue prediction values ​​for multiple categories, wherein the revenue prediction model includes multiple revenue prediction modules corresponding to the multiple categories, and each revenue prediction module is used to perform a corresponding category revenue prediction on the data feature set to obtain a corresponding category revenue prediction value; and a rating module for determining the total revenue prediction value for the corresponding store based on the revenue prediction values ​​for all categories, and determining the value rating of each store on the e-commerce platform based on the total revenue prediction value for all stores.

[0015] A third aspect of this application provides an electronic device, including: a memory and a processor, wherein the processor executes computer-readable instructions stored in the memory to implement the store rating method.

[0016] The store tier classification method provided in this application acquires store datasets from multiple stores on an e-commerce platform and constructs a data feature set. This comprehensively collects multi-dimensional information about stores and constructs features, providing a data foundation for revenue prediction. The data feature set is input into a revenue prediction model containing multiple category revenue prediction modules. Each module accurately captures key factors affecting revenue for different categories. The combined predicted values ​​from multiple categories comprehensively reflect revenue composition and trends, improving the accuracy and comprehensiveness of revenue prediction. The total predicted revenue for each store is determined based on the predicted revenue values ​​from all categories, comprehensively considering the overall revenue potential of the store while ensuring comparability of revenue situations across different stores. Classifying stores into value tiers based on the total predicted revenue for all stores helps e-commerce platforms develop differentiated operational strategies for stores of different tiers, achieving rational resource allocation and improving utilization efficiency. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram illustrating the application environment of the store rating method provided in one embodiment of this application.

[0019] Figure 2 This is a flowchart illustrating a training method for an income prediction model provided in an embodiment of this application.

[0020] Figure 3 This is a flowchart illustrating a method for determining the pseudo-Gini coefficient according to an embodiment of this application.

[0021] Figure 4 An example diagram showing the first cumulative actual percentage and the second cumulative actual percentage provided in an embodiment of this application.

[0022] Figure 5 An example diagram of a first scale curve and a second scale curve provided for an embodiment of this application.

[0023] Figure 6 This is a flowchart illustrating a method for classifying store levels according to an embodiment of this application.

[0024] Figure 7 This is a flowchart illustrating a method for deploying an income forecasting model according to an embodiment of this application.

[0025] Figure 8 This is a schematic diagram of a store rating system provided in one embodiment of this application.

[0026] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this application clearer, the application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] It should be noted that in this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and drawings of this application are used to distinguish similar objects, not to describe a specific order or sequence.

[0029] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner. Unless otherwise specified, the following embodiments and features described herein can be combined with each other.

[0030] In the cross-border e-commerce sector, with the booming development of the business, the number of stores on cross-border e-commerce platforms (such as independent websites) is increasing daily, and competition among stores is becoming increasingly fierce. Against this backdrop, traditional store operation models are gradually revealing many drawbacks.

[0031] In the traditional model, cross-border e-commerce platforms often adopt a uniform approach to allocating traffic, marketing resources, and risk management strategies. This "one-size-fits-all" operating method, failing to fully consider the differentiated needs of individual stores, has led to a series of problems. From a resource allocation perspective, a large amount of resources are wasted on low-value activities, failing to achieve efficient resource utilization. In terms of marketing, marketing activities lack targeting, resulting in poor effectiveness and making it difficult to maximize the value of marketing budgets, leading to inefficient marketing. In terms of risk management, platforms can only passively respond to problems and cannot prevent them in advance, often taking measures only after risks have already caused some impact, resulting in a lag in risk management.

[0032] To address these issues and achieve refined store management, value segmentation of cross-border e-commerce stores becomes essential. The value of a cross-border e-commerce store refers to the projected revenue it will generate for the platform over a future period. Based on this projected revenue, stores can be categorized into high, medium, and low-value stores, thus achieving value segmentation. From an operational perspective, value segmentation allows operational actions to shift from generalization to precise focus, concentrating resources on high-value aspects and avoiding waste. In marketing, it aligns marketing resources precisely with the store's conversion potential and user value, preventing blind spending and improving the return on investment (ROI) of marketing budgets. From a risk management perspective, value segmentation helps the platform shift from "passively responding to problems" to "proactively preventing risks," combining the two dimensions of risk and value to achieve more precise control over stores.

[0033] However, current methods for segmenting e-commerce stores based on their value face technological bottlenecks. Existing technologies suffer from limitations in store value assessment, including simplistic methodologies, unreasonable evaluation metrics, and weak dynamic adaptability. These issues result in severely inaccurate store grading, making it difficult to serve as a scientific basis for platform resource allocation, marketing strategy development, and risk management. The key challenge in store value segmentation lies in accurately predicting the future value of stores. Only with precise predictions of future store value can the effective segmentation of high, medium, and low-value stores be guaranteed; conversely, inaccurate predictions render store value segmentation meaningless.

[0034] To address the aforementioned issues, the store tier classification method provided in this application acquires store datasets from multiple stores on an e-commerce platform and constructs a data feature set. This comprehensively collects multi-dimensional information about stores and constructs features, providing a data foundation for revenue prediction. The data feature set is input into a revenue prediction model containing multiple category revenue prediction modules. Each module accurately captures key factors influencing revenue for different categories. The combined prediction values ​​from multiple categories comprehensively reflect revenue composition and trends, improving the accuracy and comprehensiveness of revenue prediction. The total predicted revenue for each store is determined based on the predicted revenue values ​​from all categories, comprehensively considering the overall revenue potential of the store while ensuring comparability of revenue situations across different stores. Classifying stores into value tiers based on the total predicted revenue value helps e-commerce platforms develop differentiated operational strategies for stores of different tiers, achieving rational resource allocation and improving utilization efficiency.

[0035] Please see Figure 1 This is a schematic diagram illustrating the application environment of the store rating method provided in this application embodiment. Figure 1 As shown, the technical solution of this application uses the operating environment of an e-commerce platform (e-commerce platform) 10 as its application environment. The e-commerce platform 10 can be an e-commerce service platform that provides open independent site services. The e-commerce platform 10 can be used to provide merchants' goods and services to users. Among them, users can be merchant users (e.g., sellers or product providers), customer users (e.g., buyers or product users), potential users (e.g., users who are browsing but have not yet committed to purchasing, etc.).

[0036] E-commerce platform 10 can be implemented through a processing facility including a processor and memory. When instructions stored in the memory are executed, the e-commerce platform 10 performs the e-commerce functions and supporting functions described herein. The processing facility can be a server, client, network infrastructure, mobile computing platform, cloud computing platform, or other computing platform, and provides electronic connectivity and communication between electronic components of e-commerce platform 10, merchant devices 20, client devices 30, etc. E-commerce platform 10 can be implemented as cloud computing services, Software as a Service (SaaS), Infrastructure as a Service (IaaS), Platform as a Service (PaaS), Desktop as a Service (DaaS), Managed Software as a Service (MSaaS), Mobile Backend as a Service (MBaaS), Information Technology Management as a Service (ITMaaS), etc.

[0037] E-commerce platform 10 can provide a centralized system for offering merchants online resources and facilities to manage their businesses. E-commerce platform 10 can provide communication facilities 101 and associated merchant interfaces to provide electronic communication and marketing, such as using electronic messaging aggregation facilities to collect and analyze communication interactions between merchants, customers, merchant devices 20, customer devices 30, etc., to aggregate and analyze communications. For example, in a real-world scenario, when a customer makes a paid subscription in a store, communication facility 101 analyzes the revenue generated by that operation. Based on this revenue data, specific algorithms and models can be used to determine the store's value level, providing a basis for subsequent store management and resource allocation.

[0038] The data facility 102 of the e-commerce platform 10 can store information about business and merchant transactions, and can have various methods to enhance, contribute, refine, and extract data. Over time, the collected data can enable improvements to various aspects of the e-commerce platform 10. During the store ranking process, the data facility 102 can acquire the store datasets and data feature sets for each store required for the store ranking method. By collecting and organizing this data, it provides data support for subsequent revenue prediction models, ensuring the scientific validity and rationality of the store ranking.

[0039] E-commerce platform 10 can be configured with a business management engine 104 for content management, task automation, and data management to support and serve multiple online stores 103. The business management engine 104 includes basic or "core" functionalities of e-commerce platform 10 (e.g., common to most online store activities, such as cross-channel, administrator interface, merchant location, industry, product type, etc.) that can be reused across online stores 103 (e.g., functions that can be reused / modified across core functionalities). In a specific example, the business management engine 104 can be applied to the store ranking process, providing personalized support and services to stores based on their value level. For example, high-value stores can receive resource allocation, more traffic support, marketing promotion resources, and preferential policies to help them further improve their performance and market competitiveness.

[0040] In some embodiments, merchants may use e-commerce platform 10 to manage business with customers, such as implementing e-commerce experiences with customers through online store 103 (e.g., independent website store), managing business through e-commerce platform 10, and interacting with customers through communication facilities 101 of e-commerce platform 10.

[0041] Merchants can manage one or more storefronts in online store 103 through, for example, merchant device 20 (e.g., computer, etc.) and offer goods to customers through different channels (e.g., online store 103). The terms "online store 103," "storefront," and "shop" are used synonymously and refer to the online e-commerce business unit provided by the merchant through e-commerce platform 10.

[0042] Merchants can log in to administrator 105 via merchant device 20 (such as from a desktop computer or mobile device) and manage various aspects of their online store 103. For example, merchants can customize the appearance of the online store 103 through a theme system, or customize the presentation content of the product detail page (such as product titles, product details, product images, and other types of content).

[0043] Online store 103 can support a large number of independently managed stores and process a large amount of data for each store daily. This may include, for example, basic store data, customer contact information, information about purchased goods, information about presented services, and any other information related to business conducted through e-commerce platform 10. Examples include multiple categories of revenue data, such as subscription revenue data, commission revenue data, and payment rebate data. In some embodiments, e-commerce platform 10 may store this data in data facility 102.

[0044] In some embodiments, customers can interact via customer device 30 (e.g., computer, mobile terminal, etc.) or any other business interface device known in the art. E-commerce platform 10 enables merchants to contact customers through online store 103, facilitate business with customers through dialogue via electronic communication facility 101, etc., thereby providing a system for contacting customers and facilitating merchant services for real or virtual paths that can be used to contact and interact with customers.

[0045] Online store 103 can be provided to customers by e-commerce platform 10, where customers can browse and purchase various available goods and conduct various transactions related to online store 103. For example, online store 103 can be made available to customer device 30 through a webpage provided by the server of e-commerce platform 10. The server can receive requests for webpages from a browser or other application installed on customer device 30, and in response, the server sends back the requested webpage, which is then displayed by the web browser (or other application) of customer device 30. For example, the subscription interface of online store 103 can display controls corresponding to paid subscription operations for users, facilitating user operation and providing a means for the platform to collect user behavior data and revenue data, further supporting the implementation of business functions such as store level classification.

[0046] The following will refer to the computer program product running on an e-commerce platform (such as...). Figure 1 Let's take an e-commerce platform (10) as an example to illustrate.

[0047] In one example, the store rating method provided in this application embodiment is implemented by applying a revenue prediction model. The revenue prediction model and its training method will be introduced next.

[0048] In one embodiment of this application, given the continuous nature of store revenue, a regression model can be constructed as the revenue prediction model to accurately predict store revenue. However, considering the significant differences in the magnitude of revenue across multiple revenue categories for different stores, using a uniform model for prediction would make it difficult to guarantee the accuracy of revenue predictions for each category. For example, multiple categories include subscription revenue, commission revenue, and payment rebate revenue, and the revenue data for these categories includes subscription revenue data, commission revenue data, and payment rebate data. For each of these revenue categories, multiple corresponding revenue regression prediction modules can be independently constructed.

[0049] In one example, the revenue prediction model includes multiple revenue prediction modules corresponding to multiple categories. For instance, the revenue prediction model includes a revenue prediction module for the subscription revenue category, a revenue prediction module for the commission revenue category, and a revenue prediction module for the payment rebate category. Each revenue prediction module is used to perform a corresponding category revenue prediction on the data feature set to obtain the corresponding category's revenue prediction value.

[0050] Based on the above embodiments, each module of the income prediction model focuses on predicting and analyzing income of a specific category on the data feature set, thereby outputting the income prediction value corresponding to that category, thus improving the accuracy and reliability of the prediction.

[0051] In one example, reference Figure 2 The diagram shown is a flowchart illustrating a training method for an income prediction model provided in an embodiment of this application, which includes the following steps.

[0052] Step S201: Collect the historical store dataset for each store, and construct the historical data feature set for each store based on the historical store dataset.

[0053] In one embodiment of this application, the historical store dataset includes historical basic data of the stores, which will be referred to as basic data below for ease of description. In one example, the stores here refer to the stores used to collect model training sample data, which may include stores that need to be classified into different levels later, or other stores, in order to obtain a richer dataset to train the model and improve the model's generalization ability.

[0054] Basic store data may include store ID, store registration duration, registration region, and business category. For example, if store ID A001 was registered on January 1, 2020, then as of the current statistical time, its registration duration is X years and X months; the registration region may be a province in China, such as Guangdong Province; and the business category may be the women's clothing sub-segment under the apparel category.

[0055] Based on the above embodiments, the store's basic data can provide a foundation for a comprehensive understanding of the store's background and assist in predicting revenue for a predetermined period of time. For example, different regions and categories have different market potentials, which may have a potential impact on revenue for a predetermined period of time.

[0056] In one example, the preset future time period can be 6 months, 9 months, 12 months, etc., and can be set according to actual needs. The following example will illustrate this with a preset future time period of 6 months.

[0057] In one embodiment of this application, the historical store dataset also includes historical revenue data in multiple categories. For example, these categories include subscription revenue, commission revenue, and payment rebate data. The historical revenue data in these categories includes historical subscription revenue data (hereinafter referred to as subscription revenue data), historical commission revenue data (hereinafter referred to as commission revenue data), and historical payment rebate data (hereinafter referred to as payment rebate data). This data is a key basis for subsequent model predictions of revenue over a predetermined time period.

[0058] In one example, the historical revenue data for each category in the store dataset can include data from multiple time points. Selecting multiple time points allows data to be collected from different time dimensions, enabling the model to more accurately predict the store's revenue for the next six months. The number and specific dates of the time points can be set according to actual needs. For example, multiple time points can include 12 time points, such as the 1st or 15th of each month within the most recent year.

[0059] In one example, subscription revenue data includes revenue generated by stores through paid subscriptions to platform services (such as smart marketing and one-click store decoration) under the cross-border e-commerce independent website model.

[0060] For example, subscription revenue data includes, but is not limited to, one or more of the following: the store's current subscription package type, such as monthly membership, quarterly membership, annual membership, etc.; the start and end dates of the current subscription package, such as an annual membership package starting on January 1, 2024, and ending on January 1, 2025; the store's continuous subscription duration, such as if the store has continuously subscribed since registration without interruption, then the continuous subscription duration is the duration from the registration date to the current statistical date; the store's subscription revenue for different time periods in the past, such as 1, 2, 3, 4, 5, 6, and 12 months (e.g., subscription revenue of 100 yuan in the past month, subscription revenue of 300 yuan in the past 3 months, etc.); the number of subscription months in the past 12 months, such as if the store has been subscribed for 10 months in the past 12 months, then the number of subscription months is 10; the number of unsubscribed months, continuing from the above example, the number of unsubscribed months is 2; subscription package upgrades... The number of upgrade levels is calculated as follows: for example, if a store upgrades from a monthly membership to a quarterly membership, and then to an annual membership, the upgrade count is 2 times; the maximum number of days a subscription can be interrupted is calculated as follows: if a subscription is interrupted for a period of time, the maximum interruption period is 15 days; the number of days since the last subscription ended is calculated as follows: for example, if the last subscription ended on December 31, 2023, and the current date is May 10, 2024, then the number of days since the last subscription ended is 131 days; the number of days since the start of the current subscription package is calculated as follows: for example, if the current annual membership package starts on January 1, 2024, and ends on May 10, 2024, then the number of days since the start of the current subscription is 131 days; the number of days until the end of the current subscription is calculated as follows: for example, if there are 236 days until January 1, 2025, then the number of days until the end of the current subscription is 236 days; and the remaining income after the end of the current subscription package is calculated as follows: for example, if the annual membership subscription fee is 1200 yuan, amortized monthly at 100 yuan per month, and there are 8 months remaining, then the remaining income is 800 yuan, etc.

[0061] Based on the above embodiments, subscription revenue data is crucial for predicting subscription revenue for the next six months. For example, the type of subscription package a store currently uses (monthly membership, quarterly membership, annual membership, etc.) directly affects the stability and amount of future subscription revenue; the start and end times of the current subscription package can be used to estimate the duration of the subscription status within the next six months, thereby estimating subscription revenue; the length of continuous subscriptions a store uses reflects its loyalty and dependence on the platform services, and stores with longer continuous subscription periods are more likely to generate subscription revenue in the next six months.

[0062] In one example, commission revenue data includes the revenue information obtained by the platform from taking a commission on each completed order of the store. The commission revenue data is related to the store's order conversion rate. Different membership levels (monthly members, quarterly members, annual members) correspond to different commission rates. For example, the rate for monthly members is 2%, for quarterly members it is 1%, and for annual members it is 0.5%. This determines the base for commission revenue in the next 6 months.

[0063] For example, commission revenue data includes store traffic data, which reflects the attractiveness of the store's products and market performance. Traffic volume affects the likelihood of orders being placed in the next six months, thus impacting commission revenue. Examples include store PV (page views) and UV (unique visitors) for the past 1, 3, 5, 7, 15 days, 2, 4, and 6 months; month-on-month growth rates of PV and UV for the past 7, 15, and 30 days; and month-on-month growth rates for the past 2, 4, and 6 months.

[0064] Commission revenue data also includes store product data, which reflects product pricing strategies and market positioning. Product data influences consumer purchasing decisions, thus affecting commission revenue for the next six months. For example, data for the most recent 1, 3, 5, 7, 15 days, 2, 4, and 6 months includes: the number of products listed, delisted, and penalized; product views, add-to-cart, and favorites; view conversion rate; add-to-cart conversion rate; favorites conversion rate; average price of listed products; average discount rate; maximum discount rate; minimum discount rate; percentage of products with free shipping; and the percentage of shipping costs in the product price.

[0065] Commission revenue data also includes store transaction data, used to display changes in store transaction activity and business scale. Transaction activity directly affects commission revenue for the next 6 months. For example, data for the most recent 1, 3, 5, 7, 15 days, 2, 4, and 6 months includes: number of transaction orders, transaction amount, average transaction amount, number of products traded, average daily transaction amount, average monthly transaction amount, maximum single-day transaction amount, month-on-month growth rate of transaction amount for the most recent 7, 15, and 30 days, and month-on-month growth rate of transaction amount for the most recent 2, 4, and 6 months.

[0066] Commission revenue data also includes: commission revenue for the most recent 1, 3, 5, 7, 15 days, 2, 4, and 6 months; month-on-month growth rates for commission revenue for the most recent 7, 15, and 30 days; and month-on-month growth rates for commission revenue for the most recent 2, 4, and 6 months. This data directly reflects the dynamic changes in store commission revenue and provides a reference for predicting commission revenue for the next 6 months.

[0067] Based on the above embodiments, the commission revenue data includes the revenue information obtained by the platform from each transaction order of the store. Different membership levels correspond to different commission rates, which determines the potential scale of commission revenue in the next 6 months.

[0068] In one example, the payment rebate data includes the rebate income generated by the third-party payment channel used for each successful order transaction in the store. The payment rebate data is related to the payment channel corresponding to the completed orders in the store.

[0069] For example, payment rebate data includes corresponding data on payment channels and rebate ratios, such as the rebate ratio for payment channel 1, the rebate ratio for payment channel 2, ..., the rebate ratio for payment channel n. Different rebate ratios directly affect the rebate income of payment channels in the next 6 months.

[0070] Payment rebate data also includes order payment data, which reflects the store's transaction status and payment trends through different payment channels. Transaction status and trends affect payment channel rebate revenue for the next 6 months. For example, the store's payment amount for the most recent 1, 3, 5, 7, 15 days, 2, 4, and 6 months; the top 1, 2, and 3 payment amounts, payment percentages, payment channel names, and the month-on-month growth rate of the store's payment amount for the most recent 1, 2, 4, and 6 months.

[0071] The payment rebate data also includes: the store's payment channel rebate revenue for the most recent 1, 3, 5, 7, 15 days, 2, 4, and 6 months, as well as the month-on-month growth rate of payment channel rebate revenue for the most recent 1, 2, 4, and 6 months. This data can show the dynamic changes in payment channel rebate revenue and help predict payment channel rebate revenue for the next 6 months.

[0072] Based on the above embodiments, the payment rebate data includes the rebate income information generated by the third-party payment channel used for each successful order transaction in the store. The rebate ratio and payment amount of the payment channel directly affect the amount of rebate income from the payment channel in the next 6 months.

[0073] In one embodiment of this application, obtaining a data feature set for each store based on a store dataset includes: performing data cleaning and format standardization on the store dataset to obtain a preprocessed store dataset. Based on the preprocessed store dataset, determining multiple data features for each store, including basic features corresponding to basic data, subscription revenue features corresponding to subscription revenue data, commission revenue features corresponding to commission revenue data, and payment rebate features corresponding to payment rebate data. A data feature set is then constructed based on these multiple data features.

[0074] In one example, data cleaning involves removing noise, errors, and inconsistencies from the data, such as removing duplicate order records and correcting erroneous monetary data, thereby ensuring the accuracy and integrity of the data. Format standardization involves conforming the data to a uniform format specification, such as standardizing date formats to "YYYY-MM-DD" and uniformly retaining two decimal places for monetary data, thus obtaining a preprocessed store dataset.

[0075] In one example, based on the preprocessed historical store dataset, each store has multiple data features. These multiple data features are divided according to different data types, including: historical basic features corresponding to historical basic data (referred to as basic features), historical subscription revenue features corresponding to historical subscription revenue data (referred to as subscription revenue features), historical commission revenue features corresponding to historical commission revenue data (referred to as commission revenue features), and historical payment rebate features corresponding to historical payment rebate data (referred to as payment rebate features).

[0076] Each store's various data features are selected and designed specifically to predict its revenue over the next six months. Among these, subscription revenue features, such as the store's continuous subscription duration, reflect the store's loyalty and reliance on the platform's services; stores with longer continuous subscription periods are more likely to generate subscription revenue in the next six months. The product browsing conversion rate feature in commission revenue reflects the attractiveness and sales efficiency of the store's products; stores with high conversion rates are likely to perform better in transaction commission revenue over the next six months. The payment rebate feature, including the rebate ratio and payment amount percentage of payment channels, directly affects the amount of payment channel rebate revenue in the next six months. These features collectively constitute a comprehensive description of the store's value and play a crucial role in predicting the store's revenue over the next six months.

[0077] In one example, in the fields of machine learning and data analysis, "features" refer to variables used to describe the attributes of data objects. These variables can reflect certain essential characteristics of the data objects and have a significant impact on the model's prediction results. In the scenario of cross-border e-commerce independent website store value clustering provided in the embodiments of this application, the various indicators in the listed basic data, subscription revenue data, commission revenue data, and payment rebate data are not simply raw data collections, but have been carefully selected and designed to characterize the store's operating status, user behavior, and market performance from different perspectives. They play a key role in predicting the store's revenue for a predetermined period of time (e.g., 6 months), and therefore can be called "features".

[0078] In another example, historical data feature sets can be further processed to improve model performance and prediction accuracy. For instance, further processing methods can include: feature encoding, such as converting categorical features (e.g., registration region, business category, payment channel name) into numerical form using a pre-defined encoding method so the model can process them; pre-defined encoding methods include, but are not limited to, one-hot encoding and label encoding. Feature selection, such as filtering from numerous features to identify the most valuable features for model prediction, removing redundant and irrelevant features, reducing model complexity, and improving generalization ability; feature selection can be performed using methods such as correlation analysis and feature importance assessment. Feature construction, such as constructing new features based on business knowledge and data analysis to better capture patterns and regularities in the data; for example, constructing an average transaction amount feature based on store transaction amount and order quantity, or constructing a subscription remaining duration feature based on subscription start and end times.

[0079] Based on the above embodiments, by acquiring historical store datasets from e-commerce platforms and constructing historical data feature sets, and after data cleaning and format standardization, features such as basic data, subscription revenue, commission revenue, and payment rebates are extracted. Further processing, such as feature encoding, selection, and construction, can comprehensively describe the value of stores, provide rich information for the model to accurately predict store revenue, and improve model performance and prediction accuracy.

[0080] Step S202: Based on the historical store dataset and the corresponding historical data feature set, construct a sample dataset and divide the sample dataset into a training set and a validation set.

[0081] In one example, based on the historical store dataset and historical data feature set, store sample data consisting of store number + historical sample time point + historical data features can be obtained. To accurately predict the store's revenue for the next 6 months, three types of revenue (subscription revenue, commission revenue, and payment rebate revenue) generated by the store within 6 months after each sample time point can be calculated as the target of the model prediction. The revenue data for the 6 months after each time point can be statistically calculated using a sliding time window. For example, the window length can be set to 6 months to statistically analyze the revenue data within the window after the sample time point. For example, for the store sample at the sample time point of January 10, 2024, the subscription revenue, transaction commission revenue, and payment channel rebate revenue generated in the 6 months from January 11, 2024 to July 10, 2024 can be calculated.

[0082] In one example, if the data collection time points are the 12 time points corresponding to the 12 months of the most recent year, a combination of a sliding time window and a dynamic forecast period can be used to collect income data. Specifically, for these 12 time points, starting from the first 6 time points in the time series, the income data for the next 6 months corresponding to each time point is calculated.

[0083] When processing the 7th time node out of the 12 time nodes arranged chronologically, since the remaining predictable duration is less than 6 months, the revenue data for the next 5 months corresponding to this time node can be calculated. Similarly, for the 8th time node, the revenue data for the next 4 months is calculated; for the 9th time node, the revenue data for the next 3 months is calculated; for the 10th time node, the revenue data for the next 2 months is calculated; and for the 11th time node, the revenue data for the next 1 month is calculated. However, for the 12th time node out of the 12 time nodes arranged chronologically, since it is close to the end of the data collection period and no longer has sufficient future duration for meaningful revenue prediction, its corresponding future revenue data is no longer calculated.

[0084] Thus, although it is difficult to obtain precise revenue data for the next six months for each of these 12 time points, this method of dynamically adjusting the forecast period still allows us to obtain a series of revenue forecast data with time-series characteristics. This data not only covers revenue forecasts across different time spans but also preserves the continuity and trends of the time series, providing rich training samples for subsequent revenue forecasting models.

[0085] In one example, when processing subscription revenue forecasts, the impact of different subscription types and expiration dates can be considered. Assuming that at the sample point in time, the store has an annual subscription with an expiration date more than 6 months away, then the store's subscription revenue for the next 6 months will be 50% of the guaranteed annual subscription revenue. If the store's annual subscription expires in less than 6 months, for example, 4 months away, then the subscription revenue for the next 6 months will consist of 4 months of guaranteed revenue plus 2 months of projected revenue. Similarly, if at the sample point in time the store has quarterly subscriptions with an expiration date of 2 months, then the subscription revenue for the next 6 months will be 2 months of guaranteed revenue plus 4 months of projected revenue.

[0086] In one example, the following data sequence can be constructed: store number + sample time point + historical store dataset corresponding to the store + historical data feature set corresponding to the store + three types of revenue (subscription revenue, transaction commission revenue, and payment channel rebate revenue) generated within 6 months (or other time length) after the sample time point. This data sequence is used as a sample data set. Based on the sample data of all stores at all time points, a sample dataset is obtained.

[0087] In one example, the sample dataset can be divided into a training set and a validation set according to a preset ratio. For example, the ratio could be 7:3 or 8:2. The training set is used to train the model, and the validation set is used to verify whether the model's performance meets requirements, such as whether the prediction accuracy exceeds a preset accuracy threshold (e.g., 0.8). In other examples, the sample dataset can also be divided into a training set, a validation set, and a test set, for example, in a 7:2:1 ratio.

[0088] Step S203: Using the data features of each category in the training set, train the income prediction module corresponding to the corresponding category in the income prediction model.

[0089] In one example, the data features of each category in the training set can be used to train the corresponding income prediction module in the income prediction model. The following example will illustrate how to train the income prediction module (regression model) using historical commission income features.

[0090] In one embodiment of this application, an income prediction module corresponding to the commission income category in the income prediction model is trained using historical commission income features from the training set. This includes: constructing a regression model based on the LightGBM algorithm; determining the importance score of each feature in the historical commission income features during the training process of the regression model; and determining the core features in the historical commission income features based on the importance scores. The regression model is then trained based on the core features, enabling the regression model to learn the non-linear mapping relationship between the core features and the income of the commission income category, thereby obtaining the trained income prediction module corresponding to the commission income category.

[0091] In one example, when initializing the regression model of the LightGBM algorithm, the `objective` parameter can be set to "regression". This setting indicates that the model's task type is a regression problem, aiming to predict continuous commission income, thus guiding the subsequent training and optimization of the model. Simultaneously, based on the actual data and model performance requirements, other important parameters should be configured appropriately, such as `num_leaves` (controls the maximum number of leaf nodes in the tree, affecting model complexity), `min_data_in_leaf` (the minimum amount of data required for leaf nodes to prevent overfitting), and `learning_rate` (the learning rate, controlling the step size of model parameter updates in each iteration). Properly setting these parameters can balance model complexity and generalization ability, laying the foundation for building the regression model.

[0092] In one example, a key step in training the regression model is to delve into the intrinsic relationship between historical commission income features and commission income itself. The LightGBM algorithm possesses powerful feature importance assessment capabilities, outputting an importance score for each feature within the historical commission income features based on the model's performance during training. These importance scores reflect the contribution of each feature to commission income prediction, providing a quantitative basis for selecting core features.

[0093] For example, you can view the score of each feature to understand which features play a key role in the model's prediction. A higher score indicates a greater impact of the corresponding feature. Features with scores above a preset importance score threshold are selected as core features, where the importance score threshold can be set according to actual needs.

[0094] For example, to more systematically select core features, a preset number of features with the highest importance scores can be selected as core features. The preset number can consider several factors, such as the total number of features, data complexity, and the model's sensitivity to the number of features. For instance, the preset number can be set to 3, 5, or 7, ensuring that it is less than the total number of all features in the historical commission income feature set, thus guaranteeing the representativeness and effectiveness of the selected core features.

[0095] Based on the above embodiments, by filtering core features, not only can features that contribute little to commission income prediction or even introduce noise be removed, but the input dimension of the model can also be reduced, the complexity of the model can be reduced, and the training efficiency and generalization ability of the model can be improved.

[0096] In one example, after identifying the core features, the regression model can be retrained using these features as input and commission revenue for the next six months as output. During training, the LightGBM algorithm learns the non-linear mapping relationship between the core features and commission revenue through continuous iterative optimization. This non-linear mapping relationship can capture complex patterns and regularities in the data, enabling the model to more accurately predict commission revenue under different combinations of core features.

[0097] In one example, to ensure the quality and stability of model training, appropriate training strategies and optimization algorithms can be selected. For instance, cross-validation can be used to evaluate model performance and avoid overfitting or underfitting. Simultaneously, model parameters, such as the learning rate and number of iterations, can be adjusted in a timely manner based on changes in the loss function during training to obtain the optimal combination of model parameters.

[0098] In one example, after each batch of model training is completed, the trained commission revenue prediction module can be evaluated. The training objective of the model includes minimizing the numerical difference between the predicted and actual revenue. For example, various evaluation metrics can be used, such as mean squared error, mean absolute error, and coefficient of determination, to measure the model's prediction accuracy and goodness of fit from different perspectives. By comparing and analyzing the model with actual commission revenue data, we can understand its prediction performance under different conditions.

[0099] In one embodiment of this application, the aforementioned evaluation metrics place high demands on model accuracy, making the model prone to overfitting. To address the overfitting problem, a pseudo-Gini coefficient corresponding to the income prediction module can be determined. If the pseudo-Gini coefficient is greater than a preset threshold, the income prediction module is considered to have completed training. For example, a method for determining the pseudo-Gini coefficient corresponding to the income prediction module can refer to... Figure 3 Explanation of the flowchart shown.

[0100] In one example, reference Figure 3 The diagram shown is a flowchart illustrating a method for determining the pseudo-Gini coefficient corresponding to an income forecasting module according to an embodiment of this application, which specifically includes the following steps.

[0101] Step S301: Sort the multiple stores in descending order of the revenue forecast values ​​predicted by the revenue forecast module.

[0102] In one example, based on the revenue forecast values ​​for each store predicted by the revenue forecasting module, multiple stores are sorted in descending order to obtain a store ranking table based on predicted revenue values. This ranking table can provide an ordered sequence of stores as the basis for subsequent calculations of the cumulative actual revenue percentage.

[0103] Step S302: Based on the actual revenue value of each store after sorting, determine the cumulative actual revenue percentage corresponding to multiple quantile positions in the sorted sequence, and obtain the first cumulative actual percentage.

[0104] In one example, based on the actual revenue value of each store after the above sorting, the cumulative actual revenue percentage corresponding to different quantile positions (such as 0%, 10%, 20%...100%) in the sorted sequence is calculated.

[0105] For example, for each quantile, the total actual revenue of all stores included from the start of the sequence to that quantile is calculated, and then divided by the total actual revenue of all stores to obtain the cumulative actual revenue percentage corresponding to that quantile. These percentage data are then summarized to obtain the first cumulative actual revenue percentage.

[0106] For example Figure 4 As shown, when the quantile is 10%, the total actual revenue of the top 10% of stores is calculated, and its proportion of the total actual revenue of all stores is calculated. This process is repeated until the 100% quantile is reached.

[0107] Step S303: Sort multiple stores in a random order, and determine the cumulative actual revenue percentage corresponding to multiple quantile positions in the random sorting sequence based on the actual revenue value of each store after random sorting, so as to obtain the second cumulative actual percentage.

[0108] In one example, multiple shops are reordered in a randomized sequence to form a randomized sequence. Based on the actual revenue of each shop after random sorting, the cumulative actual revenue percentage corresponding to each quantile position in the randomized sequence (the same as the quantile setting for sequential sorting) is calculated.

[0109] For example, for each quantile in a randomly sorted sequence, the total actual revenue of the stores from the beginning of the sequence to that quantile position is calculated, and then divided by the total actual revenue of all stores to obtain the cumulative actual revenue percentage corresponding to that quantile position, and then the second cumulative actual revenue percentage is obtained.

[0110] For example Figure 4 As shown, under random sorting, theoretically the proportion of real income within each quantile interval should be evenly distributed, for example, the proportion within each tenth percentile is 10%, and the cumulative proportion increases monotonically by 10%.

[0111] Step S304: Draw the first proportion curve based on the first cumulative actual proportion.

[0112] In one example, a first proportion curve is plotted based on the first cumulative actual percentage. The horizontal axis of the first proportion curve indicates the quantile, and the vertical axis indicates the first cumulative actual percentage. The quantile, used as the horizontal axis, ranges from 0% to 100%; the first cumulative actual percentage, used as the vertical axis, creates a first proportion curve that visually reflects the changing trend of the actual cumulative percentage of income ranked based on the model's predicted income. For example... Figure 5 As shown, the first proportional curve illustrates the relationship between the model's predicted income ranking and the cumulative proportion of actual income.

[0113] Step S305: Draw the second proportion curve based on the second cumulative actual proportion.

[0114] In one example, a second proportional curve is plotted based on the second cumulative actual percentage. The horizontal axis of the second proportional curve indicates the quantile, and the vertical axis indicates the second cumulative actual percentage. Similar to the first proportional curve, the horizontal axis is plotted with the quantile, ranging from 0% to 100%; the vertical axis is plotted with the second cumulative actual percentage, representing the trend of the changing proportion of real income under random ranking. For example... Figure 5 As shown, the second proportion curve reflects the distribution of the cumulative proportion of real income when randomly ranked.

[0115] Step S306: Determine the area between the first proportional curve and the second proportional curve, and determine the pseudo-Gini coefficient based on the area.

[0116] In one example, the area between the first and second scale curves is determined, for example, by calculating the area between the two curves using methods such as numerical integration. This area is a key metric for measuring the difference between the model's predicted income ranking and the actual income ranking.

[0117] In one example, the pseudo-Gini coefficient is determined based on the area. For instance, the formula for calculating the pseudo-Gini coefficient can be expressed as: Pseudo-Gini coefficient = 2 × (Area - 0.5). The specific value of the pseudo-Gini coefficient can be obtained by substituting the area between the two curves calculated above into the formula.

[0118] In one example, the pseudo-Gini coefficient typically ranges from [0, 1]. If the pseudo-Gini is 1, the model's predicted revenue ranking is completely consistent with the actual revenue ranking, indicating that the model has reached its optimal ranking capability and can accurately rank stores according to their actual revenue. If the pseudo-Gini is 0, the model's predicted revenue ranking is indistinguishable from random ranking, meaning the model's predictions provide no useful information about the actual revenue ranking, indicating extremely poor ranking capability. If the pseudo-Gini is less than 0, the model's predicted revenue ranking is inversely related to the actual revenue ranking; that is, the model predicts stores with high actual revenue as low revenue and vice versa, a situation undesirable in actual business. In general business applications, a pseudo-Gini exceeding a preset threshold (e.g., 0.60) indicates a relatively excellent ranking capability, accurately reflecting the actual revenue ranking of stores to a certain extent and providing valuable reference for business decisions.

[0119] Based on the above embodiments, to address the issue of overfitting caused by high accuracy requirements of various evaluation indicators, a pseudo-Gini coefficient corresponding to the revenue prediction module is determined. Specifically, stores are sorted according to their predicted revenue values, and the cumulative actual revenue percentages at different quantiles under sequential and random sorting are calculated. Then, a first proportional curve and a second proportional curve are plotted, and the area between the first and second proportional curves is calculated to determine the pseudo-Gini coefficient. When the pseudo-Gini coefficient is greater than a preset threshold, it indicates that the model's predicted revenue ranking is close to the actual ranking, demonstrating excellent model ranking ability, and the revenue prediction module can be considered successfully trained. Thus, it is only necessary to focus on whether the ranking of predicted revenue is consistent with the ranking of actual revenue, rather than whether the magnitude of the predicted revenue value is similar to the actual revenue value. This reduces the predictive pressure on the revenue regression prediction model, provides valuable reference for business decisions, effectively alleviates the fitting problem of over-regression prediction models, and ensures the model's practicality.

[0120] In one embodiment of this application, after the income prediction module has been trained, the updated pseudo-Gini coefficient corresponding to the income prediction module can be determined based on the validation set. If the updated pseudo-Gini coefficient is less than or equal to a preset coefficient threshold, the income prediction module is retrained.

[0121] Based on the above embodiments, once the revenue prediction module has completed its initial training, its performance can be further evaluated using a validation set to ensure that the module has reliable and excellent ranking prediction capabilities in practical applications.

[0122] In another example, the regression prediction model can also include only one regression prediction module to predict the store's total revenue. This makes subsequent model deployment and application easier, especially when stores on other e-commerce platforms have multiple revenue streams.

[0123] In another example, besides using a regression model to predict specific revenue figures, one can first categorize stores into high, medium, and low value levels based on their actual revenue, and then label the store samples as high, medium, and low, thus constructing a three-class classification model. Compared to regression models, three-class classification models are less difficult to predict; however, because changes in the overall revenue level of the platform lead to changes in the proportion of high, medium, and low value stores, it may be necessary to increase the frequency of model iteration and updates.

[0124] In one example, the above embodiment introduced a method for training a revenue prediction model. The following section will introduce a method for applying this revenue prediction model to classify stores into different tiers. Please refer to [link / reference]. Figure 6 The diagram shown is a flowchart illustrating the store rating classification method provided in an embodiment of this application. In one embodiment of this application, the method includes the following steps: Step S601: Obtain the store dataset for each store among multiple stores on the e-commerce platform, and construct the data feature set for each store based on the store dataset.

[0125] In one example, the "stores" here refer to stores that need to be categorized into different levels. The method for collecting the store dataset and the data within it can be found in the description of the embodiment in step S201, and will not be repeated here. For example, the store dataset includes basic store data, subscription revenue data, commission revenue data, and payment rebate data.

[0126] In one example, the method for constructing a data feature set for each store based on the store dataset, and the data in the data feature set, can be referred to the description in the embodiment of step S201, and will not be repeated here. For example, the data feature set includes basic features, subscription revenue features, commission revenue features, and payment rebate features.

[0127] Step S602: Input the data feature set into the preset income prediction model, and use the income prediction model to predict the income of the data feature set to obtain income prediction values ​​for multiple categories.

[0128] In one embodiment of this application, the revenue prediction model represents a pre-trained regression model. This regression model can learn the nonlinear mapping relationship between features in the data feature set and store revenue, thereby enabling revenue prediction based on the input data feature set to obtain the predicted revenue of the store in the future for a preset time period.

[0129] In one example, the revenue prediction model includes multiple revenue prediction modules corresponding to multiple categories. For instance, the model might include modules for subscription revenue, commission revenue, and payment rebates. Each module is used to predict the revenue for a specific category from the data feature set, resulting in a predicted revenue value for that category. The training method for this revenue prediction model can be found in [reference needed]. Figure 2 Description of the illustrated embodiment.

[0130] Based on the above embodiments, the data feature set is input into a preset revenue prediction model. This model, as a pre-trained regression model, can deeply learn the non-linear mapping relationship between data features and store revenue, thereby accurately predicting store revenue within a preset time period. The revenue prediction model includes multiple revenue prediction modules corresponding to different categories. Each module performs predictions for different revenue categories such as subscriptions, commissions, and payment rebates, effectively improving the targeting and accuracy of the predictions and providing strong support for comprehensive and detailed prediction of store revenue.

[0131] Step S603: Based on the revenue forecast values ​​of all categories, determine the total revenue forecast value of the corresponding store, and determine the value level of each store in the e-commerce platform based on the total revenue forecast value of all stores.

[0132] In one embodiment of this application, after forecasting various revenue categories and obtaining predicted revenue values ​​for different categories such as subscription revenue, commission revenue, and payment rebate revenue, these predicted revenue values ​​for different categories can be aggregated and calculated in order to grasp the overall revenue situation of the store over a predetermined period of time. For example, the predicted revenue values ​​for each category can be added together. For instance, the predicted revenue values ​​for the subscription revenue category, the commission revenue category, and the payment rebate category can be added together one by one, and the final sum is the total predicted revenue value for the store.

[0133] Based on the above embodiments, the total revenue forecast can comprehensively reflect the store's expected revenue based on different business segments in a specific future period. It provides key data references for the store's operational decisions, resource allocation, and business planning, and helps store managers understand the store's revenue trend from a macro perspective, thereby formulating more reasonable and effective business strategies.

[0134] In one embodiment of this application, the method for determining the revenue level of each store based on the total projected revenue can be referred to... Figure 7 The following steps are shown in the flowchart.

[0135] Step S701: Sort the multiple stores in descending order of total projected revenue.

[0136] In one example, multiple stores are sorted from highest to lowest total projected revenue to form a store value sequence based on projected revenue. This sorting provides a benchmark framework for subsequent value grading, ensuring that high-value stores are at the top of the sequence.

[0137] Step S702: Based on the total revenue forecast of each store after sorting, determine the cumulative forecast revenue percentage corresponding to multiple quantile positions in the sorting sequence.

[0138] In one example, based on the sorted sequence of stores, the cumulative predicted revenue percentage is calculated at preset quantile nodes (e.g., 0%, 1%, 2%...100%). For instance, a 1% granularity is used for fine-grained statistics on the top 10% range. The cumulative predicted revenue percentage for each quantile is obtained by summing the predicted revenue of stores from the start of the sequence to each quantile and dividing by the total predicted revenue of all stores. This accurately captures the distribution characteristics of high-value stores; for example, the top 1% quantile may correspond to more than 80% of the cumulative revenue percentage, conforming to the inverted pyramid principle.

[0139] Step S703: Plot the predicted revenue ratio curve based on the cumulative predicted revenue ratio.

[0140] In one example, the horizontal axis of the predicted revenue ratio curve indicates quantiles, and the vertical axis indicates the cumulative predicted revenue share. Plotting the predicted revenue ratio curve with quantiles on the horizontal axis and cumulative predicted revenue share on the vertical axis can intuitively reflect the central tendency of the predicted revenue distribution. The curve shape, which is steep at the beginning and flat at the end, indicates the strong contribution of high-value stores, providing a visual basis for subsequent curvature analysis.

[0141] Step S704: Determine the curvature of the predicted income ratio curve and determine the division position based on the quantile corresponding to the maximum curvature.

[0142] In one example, the curvature of the curve at each point can be calculated mathematically. For instance, the curvature of the predicted revenue proportion curve at each quantile can be determined using a second-order derivative algorithm. The quantile position corresponding to the maximum curvature is then identified, marking a key inflection point in the distribution of the total predicted revenue. For example, before this quantile position, the cumulative revenue contribution from individual stores shows an accelerating growth trend. Combining business experience with quantile patterns in the pseudo-Gini coefficient table, such as the cumulative proportion change of each 1% within the top 10% interval, a comprehensive threshold for classifying high-value stores can be determined.

[0143] Step S705: Among the sorted stores, the stores located before the division position are classified as high-value stores.

[0144] In one example, stores located before the quantile of maximum curvature in the sorted sequence are classified as high-value stores. These stores have the characteristics of having the smallest number of stores (e.g., 5%) but the largest revenue contribution (e.g., over 60%), which meets the dual constraints of the pyramidal quantity distribution and the inverted pyramidal value distribution.

[0145] Step S706: Based on the pyramid segmentation rule, stores other than high-value stores are divided into medium-value stores and low-value stores.

[0146] In one example, based on the pyramid segmentation rule, the remaining stores are further divided according to a preset ratio (e.g., 20% medium value, 75% low value). Simultaneously, the segmentation process must be validated: the average real revenue of stores at each level must be calculated to ensure that high value > medium value > low value with a significant difference, such as the average revenue of high-value stores being more than three times that of medium-value stores.

[0147] In one embodiment of this application, overfitting may occur during model training. For example, the model may perform well on training data but poorly on new data. In practical applications, its generalization ability can be tested using sample data from other time points at preset intervals. For example, after training the model with data from before January, it can be validated with data from February and March to ensure that the model's logic for classifying store value is consistent across different months. For example, the model's pseudo-Gini coefficient can be verified to be greater than 0.6, and the actual income distribution of high, medium, and low-value stores at different time points can be approximated to avoid inversion.

[0148] For example, train the model using January data to categorize stores into high, medium, and low value (e.g., 5:20:75). Validate using February data: calculate the pseudo-Gini coefficient; if it's 0.65 (meets the standard), proceed to the next step; if it's 0.55 (does not meet the standard), adjust the model. Statistically analyze the actual revenue distribution: high-value stores account for 62% of revenue (approximately the 60% in January), medium-value stores 24%, and low-value stores 14% (no inversion), thus passing validation. If the February validation fails, adjust the model parameters or reclassify the value proportions, then validate again using March data until the standard is met twice consecutively.

[0149] Based on the above embodiments, through dual verification of mathematical calculations and business experience, the scientific nature of the value level classification is ensured, as well as the stability and interpretability of the model in practical applications, ultimately forming a value level system that can guide business decisions such as risk management and marketing budget allocation.

[0150] In one embodiment of this application, after determining the value level of each store on the e-commerce platform, corresponding platform resources can be determined for multiple stores based on the value level.

[0151] In one example, platform resources can include categories such as traffic, marketing, services, and data, with significantly different allocation priorities for stores of different tiers. High-value stores (e.g., representing 5%) receive core resources such as homepage recommendations, search ranking boosts, and exclusive promotional subsidies. The goal is to solidify their leading position through traffic allocation and marketing privileges, ensuring their contribution to Gross Merchandise Volume (GMV) remains above 60%. Mid-value stores (e.g., representing 20%) are allocated resources such as category page recommendations and general coupons, with a focus on driving their migration towards higher value, for example, dynamically adjusting traffic allocation based on GMV growth rate. Low-value stores (e.g., representing 75%) can only use the remaining traffic pool and basic coupons; resource allocation prioritizes cost control to avoid ineffective investment.

[0152] In one example, resource allocation for high-value stores needs to reinforce their role as platform benchmarks. For instance, a high-value store with a monthly GMV accounting for 15% of the platform's total could be allocated top 3 search rankings, a central position in the main venue of major promotional events, and a "cost-price sales + platform subsidy" model (e.g., a product sold at cost price with the platform covering the difference), along with dedicated customer service channels and expedited refund services. The allocation strategy for mid-value stores focuses on incentivizing growth. If a store's GMV grows by more than 10% month-on-month for three consecutive months, it is upgraded to a "potential high-value store" and temporarily granted access to high-value resource packages. Resource allocation for low-value stores is primarily restrictive, such as limiting their participation in high-cost marketing activities (live-streaming sales require self-payment of slot fees), only opening access to low-cost channels like "flash sales," and disabling dedicated customer service.

[0153] In one example, to avoid resource stagnation, a dynamic rule of "monthly evaluation + quarterly adjustment" can be implemented. Evaluation indicators are set differently based on store level: high-value stores focus on GMV share and user repurchase rate; medium-value stores focus on GMV growth rate and traffic conversion rate; low-value stores focus on absolute GMV and return rate. Adjustment rules include upgrading (medium-value stores that meet high-value standards for two consecutive months are upgraded), downgrading (high-value stores that fail to meet standards for three consecutive months are downgraded), and elimination (low-value stores that show no improvement for six consecutive months have their number of stores limited). For example, if a medium-value store achieves a GMV growth rate of 25% in a quarterly evaluation (exceeding the platform average by 10%), its search ranking weight will be increased by 20% in the following quarter, and its application channel for high-value activities will be opened.

[0154] In one example, resource allocation can be accompanied by risk management measures. High-value stores can sign "performance-based agreements" (such as returning part of the subsidy after achieving GMV targets) to prevent resource abuse; low-value stores can have resource usage thresholds set (such as paying a deposit to participate in marketing activities) to reduce platform losses. Simultaneously, cross-departmental collaboration is needed to optimize resource utilization efficiency: working with the supply chain department to ensure high-value store inventory priority (such as allocating inventory preferentially for iPhone launches); and collaborating with the technology department to dynamically adjust search ranking weights through reinforcement learning algorithms to improve traffic conversion rates.

[0155] For example, after implementing this solution, a beauty e-commerce platform saw a significant improvement in resource allocation efficiency. Before implementation, high-value stores (5%) contributed 50% of GMV but occupied 70% of resources, while low-value stores (75%) contributed only 10% of GMV but occupied 20% of resources. After implementation, the GMV share of high-value stores increased to 65%, while resource consumption decreased to 55%. The GMV of mid-value stores increased by 30%, and their share increased from 20% to 25%. Resource consumption of low-value stores was reduced to 10%, and their GMV share stabilized at 10%. Ultimately, the platform's traffic conversion rate increased from 1.2% to 1.8%, the marketing subsidy ROI was optimized from 1:3 to 1:5, and overall profitability was significantly enhanced.

[0156] The store tier classification method provided in this application acquires store datasets from multiple stores on an e-commerce platform and constructs a data feature set. This comprehensively collects multi-dimensional information about stores and constructs features, providing a data foundation for revenue prediction. The data feature set is input into a revenue prediction model containing multiple category revenue prediction modules. Each module accurately captures key factors affecting revenue for different categories. The combined prediction values ​​from multiple categories comprehensively reflect revenue composition and trends, improving the accuracy and comprehensiveness of revenue prediction. The total predicted revenue value for each store is determined based on the predicted revenue values ​​from all categories, comprehensively considering the overall revenue potential of the store while ensuring comparability of revenue situations across different stores. Classifying stores into value tiers based on the total predicted revenue value helps e-commerce platforms develop differentiated operational strategies for stores of different tiers, achieving rational resource allocation and improving utilization efficiency. Simultaneously, by combining risk dimensions with different control measures for stores of different tiers, the overall operational efficiency and competitiveness of the platform are effectively improved.

[0157] Figure 7This is a flowchart illustrating the method for deploying a revenue prediction model according to an embodiment of this application. The model can be deployed online, primarily using a T+1 offline deployment strategy, which includes: Feature engineering deployment: Daily batch processing of the input features from all store samples to obtain the feature data required for model inference. Model inference deployment: Daily model prediction and inference using the feature data from all store samples to obtain multiple predicted revenues for each store. Value grading strategy deployment: Calculating the total predicted revenue for each store and, based on the previously determined proportions of high, medium, and low-value stores, assigning high, medium, and low value labels to each store, and outputting this information to the business for evaluation. Monitoring deployment: Statistical analysis of the pseudo-Gini coefficient between the model's predicted revenue and actual revenue to ensure timely detection and model iteration when the model's ranking performance declines; additionally, statistical analysis of the actual revenue distribution of high, medium, and low-value stores is required, and if the distribution deviation is significant, the grading needs to be reclassified.

[0158] Based on the above embodiments, full-scale store feature engineering calculations are performed in batches every night, providing standardized input for model inference. The following day, full-scale prediction is performed based on offline features, simultaneously outputting multi-dimensional revenue prediction results and value level labels to support accurate business decision-making. The pseudo-Gini coefficient is used to monitor the model's ranking capability in real time, ensuring rapid iteration when prediction accuracy declines. Simultaneously, the revenue distribution of value-level stores is dynamically tracked, and the grading strategy is automatically adjusted when the actual distribution deviates from the preset proportion by more than a threshold, forming a complete closed loop of "prediction-application-feedback-optimization." This solution effectively solves the pain points of model decay and rigid grading in offline deployment while ensuring low system latency (T+1 cycle) and high scalability, providing reliable data-driven support for e-commerce platform resource allocation and marketing strategy formulation.

[0159] Figure 8 This is a structural diagram of a store rating device provided in one embodiment of this application.

[0160] In some embodiments, the store rating device 80 may include a plurality of functional modules composed of computer program segments. The computer programs of each program segment in the store rating device 80 may be stored in the memory of an electronic device and executed by at least one processor to perform the store rating function.

[0161] In this embodiment, the store rating device 80 can be divided into multiple functional modules according to its functions. These functional modules may include: a data processing module 801, a model prediction module 802, and a rating module 803. The term "module" in this application refers to a series of computer program segments that can be executed by at least one processor and perform a fixed function, stored in memory. In this embodiment, the functional implementation of each module in the store rating device 80 can be found in the above description of the store rating method, and will not be repeated here.

[0162] The data processing module 801 is used to obtain the store dataset of each store in multiple stores on the e-commerce platform, and construct the data feature set of each store based on the store dataset.

[0163] The model prediction module 802 is used to input the data feature set into a preset income prediction model, and use the income prediction model to predict the income of the data feature set to obtain income prediction values ​​for multiple categories. The income prediction model includes multiple income prediction modules corresponding to the multiple categories. Each income prediction module is used to perform income prediction for a corresponding category on the data feature set to obtain the income prediction value for the corresponding category.

[0164] The rating module 803 is used to determine the total revenue forecast of the corresponding store based on the revenue forecast values ​​of all categories, and to determine the value level of each store in the e-commerce platform based on the total revenue forecast values ​​of all stores.

[0165] Another embodiment of this application also provides an electronic device. Figure 1 The application environment described is merely an example. In other exemplary embodiments, the computer program implementing the store rating method of this application can also run on any electronic device with sufficient computing power (such as...). Figure 9 The electronic device shown executes each step of the store rating method to provide the function of store rating.

[0166] Please see Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 9 As shown, in one embodiment of this application, the electronic device 90 can be a mobile phone, tablet computer, smart wearable device, augmented reality (AR) / virtual reality (VR) device, laptop computer, netbook, etc. This application embodiment does not limit the specific type of electronic device 90.

[0167] like Figure 9As shown, the electronic device 90 may include, but is not limited to, a communication module 91, a memory 92, a processor 93, an input / output (I / O) interface 94, and a bus 95. The processor 93 is coupled to the communication module 91, the memory 92, and the I / O interface 94 via the bus 95.

[0168] Those skilled in the art will understand that the schematic diagram is merely an example of the electronic device 90 and does not constitute a limitation on the electronic device 90. It may include more or fewer components than shown, or combine certain components, or different components. For example, the electronic device 90 may also include network access devices, etc.

[0169] Communication module 91 may include a wired communication module and / or a wireless communication module. The wired communication module may provide one or more wired communication solutions such as Universal Serial Bus (USB) and Controller Area Network (CAN). The wireless communication module may provide one or more wireless communication solutions such as Wireless Fidelity (Wi-Fi), Bluetooth (BT), mobile communication networks, Frequency Modulation (FM), Near Field Communication (NFC), and Infrared (IR) technologies.

[0170] The memory 92 can be used to store computer-readable instructions and / or modules. The processor 93 implements various functions of the electronic device 90 by running or executing the computer-readable instructions and / or modules stored in the memory 92, and by calling the data stored in the memory 92. The memory 92 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device 90, etc. The memory 92 may include non-volatile and volatile memory, such as: hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other storage devices.

[0171] The memory 92 can be the external memory and / or internal memory of the electronic device 90. Furthermore, the memory 92 can be a memory in physical form, such as a memory stick, a TF card (Trans-flash Card), etc.

[0172] Processor 93 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. Processor 93 is the computational core and control center of electronic device 90, connecting various parts of the electronic device 90 through various interfaces and lines, and executing the operating system of electronic device 90, as well as various installed application programs and program code.

[0173] For example, computer-readable instructions can be divided into one or more modules / submodules / units, one or more of which are stored in memory 92 and executed by processor 93 to complete this application. One or more modules / submodules / units can be a series of computer-readable instruction segments capable of performing a specific function, which describe the execution process of the computer-readable instructions in electronic device 90. For example, computer-readable instructions can be divided into the modules described above.

[0174] If the modules / units integrated in the electronic device 90 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by instructing related hardware through computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium, and when executed by a processor, they can implement the steps of the various method embodiments described above.

[0175] Computer-readable instructions include computer-readable instruction code, which can be in the form of source code, object code, executable file, or some intermediate form. Computer-readable media can include: any entity or device capable of carrying computer-readable instruction code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), and random access memory (RAM).

[0176] In conjunction with the above embodiments, the memory 92 in the electronic device 90 stores computer-readable instructions, and the processor 93 can execute the computer-readable instructions stored in the memory 92 to implement the store level classification method as described in the above embodiments. Specifically, the specific implementation method of the above computer-readable instructions by the processor 93 can be found in the description of the relevant steps in the above embodiments, and will not be repeated here.

[0177] I / O interface 94 is used to provide a channel for user input or output. For example, I / O interface 94 can be used to connect various input and output devices, such as mouse, keyboard, touch device, display screen, etc., so that users can enter information or visualize information.

[0178] Bus 95 is used at least to provide a channel for communication between communication modules 91, memory 92, processor 93, and I / O interface 94 in electronic device 90.

[0179] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0180] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0181] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0182] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of this application is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within this application. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0183] Furthermore, it is clear that the word "including" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices can also be implemented by a single unit or device through software or hardware. Terms such as "first," "second," etc., are used to indicate names and do not indicate any specific order.

[0184] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of this application without departing from the spirit and scope of the technical solutions of this application.

Claims

1. A method for classifying store levels, characterized in that, The method includes: Obtain the store dataset for each store in multiple stores on an e-commerce platform, and construct a data feature set for each store based on the store dataset; The data feature set is input into a preset income prediction model, and the income prediction model is used to predict the income of the data feature set to obtain income prediction values ​​for multiple categories. The income prediction model includes multiple income prediction modules corresponding to the multiple categories. Each income prediction module is used to perform income prediction for a corresponding category of the data feature set to obtain the income prediction value for the corresponding category. The total revenue forecast for each store is determined based on the revenue forecast values ​​for all categories, and the value level of each store on the e-commerce platform is determined based on the total revenue forecast values ​​for all stores.

2. The store rating method according to claim 1, characterized in that, The store dataset includes basic store data, subscription revenue data, commission revenue data, and payment rebate data. The step of obtaining the data feature set for each store based on the store dataset includes: The store dataset is cleaned and standardized to obtain a preprocessed store dataset. Based on the preprocessed store dataset, multiple data features are determined for each store. These multiple data features include basic features corresponding to the basic data, subscription revenue features corresponding to the subscription revenue data, commission revenue features corresponding to the commission revenue data, and payment rebate features corresponding to the payment rebate data. The data feature set is constructed based on the various data features.

3. The store rating method according to claim 1, characterized in that, The multiple categories include subscription revenue categories, commission revenue categories, and payment rebate categories, and the method further includes: Collect historical store datasets for each store, and construct historical data feature sets for each store based on the historical store datasets. The historical data feature sets include historical basic features, historical subscription revenue features, historical commission revenue features, and historical payment rebate features. Based on the historical store dataset and the corresponding historical data feature set, a sample dataset is constructed, and the sample dataset is divided into a training set and a validation set. Using the data features of each category in the training set, the income prediction module corresponding to the corresponding category in the income prediction model is trained.

4. The store rating method according to claim 3, characterized in that, The method further includes: using the historical commission income features in the training set to train the income prediction module corresponding to the commission income category in the income prediction model, including: A regression model based on the LightGBM algorithm is constructed. During the training process of the regression model, the importance score of each feature in the historical commission income feature is determined, and the core feature in the historical commission income feature is determined based on the importance score. The regression model is trained based on the core features, so that the regression model learns the non-linear mapping relationship between the core features and the income of the commission income category, and obtains the income prediction module corresponding to the commission income category after training.

5. The store rating method according to claim 4, characterized in that, The method further includes: Determine the pseudo-Gini coefficient corresponding to the income prediction module. If the pseudo-Gini coefficient is greater than a preset coefficient threshold, determine that the income prediction module has completed training.

6. The store rating method according to claim 5, characterized in that, Determining the pseudo-Gini coefficient corresponding to the income forecasting module includes: The stores are sorted in descending order of the revenue forecast values ​​predicted by the revenue forecasting module. Based on the actual revenue value of each store after sorting, determine the cumulative actual revenue percentage corresponding to multiple quantile positions in the sorted sequence, and obtain the first cumulative actual percentage. The multiple stores are sorted in a random order. Based on the actual revenue value of each store after random sorting, the cumulative actual revenue percentage corresponding to multiple quantile positions in the random sorting sequence is determined to obtain the second cumulative actual percentage. A first proportion curve is plotted based on the first cumulative actual proportion, where the horizontal axis of the first proportion curve indicates the quantile and the vertical axis indicates the first cumulative actual proportion. A second proportional curve is plotted based on the second cumulative actual percentage, where the horizontal axis of the second proportional curve indicates the quantile and the vertical axis indicates the second cumulative actual percentage. The area between the first proportional curve and the second proportional curve is determined, and the pseudo-Gini coefficient is determined based on the area.

7. The store rating method according to claim 5, characterized in that, After determining that the income prediction module has completed training, the method further includes: The updated pseudo-Gini coefficient corresponding to the income prediction module is determined based on the validation set. If the updated pseudo-Gini coefficient is less than or equal to a preset coefficient threshold, the income prediction module is retrained.

8. The store rating method according to claim 1, characterized in that, The process of determining the revenue level of each store on the e-commerce platform based on the total projected revenue of all stores includes: The stores are sorted in descending order of the projected total revenue; Based on the total revenue forecast for each store after sorting, determine the cumulative forecast revenue percentage corresponding to multiple quantile positions in the sorted sequence; A forecast revenue ratio curve is plotted based on the cumulative forecast revenue ratio, where the horizontal axis of the forecast revenue ratio curve indicates the quantile and the vertical axis indicates the cumulative forecast revenue ratio. Determine the curvature of the predicted income ratio curve, and determine the division position based on the quantile corresponding to the maximum curvature; The stores that are ranked before the specified division position among the sorted stores are classified as high-value stores; Based on the pyramid segmentation rule, stores other than high-value stores are divided into medium-value stores and low-value stores.

9. The store rating method according to claim 1, characterized in that, After determining the value level of each store on the e-commerce platform, the method further includes: Based on the value level, corresponding platform resources are determined for the multiple stores.

10. An electronic device, characterized in that, include: Memory, and A processor that executes computer-readable instructions stored in the memory to implement the store rating method as described in any one of claims 1 to 9.

Images