Brand recommendation methods, devices and electronic equipment

Abstract

This application provides a brand recommendation method, device, and electronic device. The method mainly involves forcibly retaining the operating brands of low-risk stores in the recommendation list of the corresponding venue, so that the recommendation results can automatically reflect the current risk status of the stores, thereby improving the matching degree between the recommendation results and actual business needs, reducing manual adjustments, and improving the timeliness and accuracy of investment promotion decisions.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of computer, and particularly relates to a brand recommendation method and device and electronic equipment. BACKGROUND

[0002] In commercial real estate operation, brand and site matching recommendation usually relies on existing recommendation means to provide candidate brand reference for business personnel. Meanwhile, a large amount of operation data is generated in the process of store operation, which can be used to evaluate and warn the survival risk of the store.

[0003] However, the information relied on by the traditional brand recommendation focuses on the matching of the brand and the site itself, and the reflection ability of the store operation state change is limited. When the store operation state fluctuates, the original recommendation result may not be able to adapt to this change in time, and there is a deviation between the brands in the recommendation list and the actual operation demand, so the business personnel still need to rely on manual judgment for adjustment, which affects the timeliness and accuracy of the recommendation result. SUMMARY

[0004] The embodiments of the present application provide a brand recommendation method, device and electronic equipment, which can improve the timeliness and accuracy of business decision.

[0005] The first aspect of the embodiments of the present application provides a brand recommendation method, comprising: obtaining operation data of a plurality of stores, brand feature data of a plurality of candidate brands and site feature data of a plurality of sites, wherein the brand feature data is used to represent the operation attribute of the candidate brand, and the site feature data is used to represent the physical condition and operation condition of the site; calling a pre-trained risk prediction model to process the operation data of the plurality of stores, wherein the risk prediction model outputs the risk state of each store, and the risk state includes a low risk state and a high risk state; calling a pre-trained recommendation model to process the brand feature data of the plurality of candidate brands and the site feature data of the plurality of sites, wherein the recommendation model outputs a recommended brand list corresponding to each site; establishing a mapping relationship between a target site corresponding to a target store and an in-operation brand, wherein the target store is a store with a low risk state in the plurality of stores, the target site is the site where the target store is located, and the in-operation brand is the brand currently operated by the target store; performing intervention processing on the recommended brand list corresponding to the target site according to the mapping relationship, so as to replace the last brand in the recommended brand list with the in-operation brand, to obtain an updated recommended brand list, and output the updated recommended brand list.

[0006] In the technical solution of the present application, by forcibly retaining the in-operation brand of the low-risk store in the recommended list of the corresponding site, the recommended result can automatically reflect the current risk state of the store, thereby improving the matching degree of the recommended result and the actual operation demand, reducing manual adjustment, and improving the timeliness and accuracy of the marketing decision.

[0007] Optionally, in a possible implementation manner of the first aspect, the above-mentioned performing intervention processing on the recommended brand list corresponding to the target site according to the mapping relationship to replace the last brand in the recommended brand list with the in-operation brand comprises: in the case that the recommended brand list already contains the in-operation brand, skipping the operation of replacing the last brand in the recommended brand list with the in-operation brand; in the case that the recommended brand list does not contain the in-operation brand, replacing the last brand in the recommended brand list with the in-operation brand. In this way, by skipping the replacement when the recommended list already contains the in-operation brand, unnecessary repeated intervention on the existing result is avoided, thereby improving the execution efficiency of the intervention processing and ensuring the stability of the recommended list.

[0008] Optionally, in another possible implementation manner of the first aspect, the above-mentioned establishing the mapping relationship between the target site corresponding to the target store and the in-operation brand comprises: screening stores with a low-risk state from a plurality of stores to obtain a target store set; batch acquiring the site identifier and the brand identifier currently operated by each store in the target store set by calling a preset external interface; and establishing the mapping relationship between each target site and the corresponding in-operation brand according to the site identifier and the brand identifier. In this way, by batch acquiring the site identifier and the brand identifier by calling the external interface and establishing the mapping relationship, the manual query link is reduced, thereby improving the efficiency and accuracy of the mapping relationship construction, and enabling the risk early warning result to timely drive the subsequent recommended intervention.

[0009] Optionally, in a further possible implementation manner of the first aspect, the calling the pre-trained recommendation model to process the brand feature data of the plurality of candidate brands and the site feature data of the plurality of sites, and outputting the recommended brand list corresponding to each site by the recommendation model comprises: for each site, in a case where the site has an operating category identifier and there is a candidate brand in the plurality of candidate brands that matches the operating category identifier, determining the candidate brand that matches the operating category identifier as a candidate brand set of the site; in a case where the site does not have an operating category identifier or there is no candidate brand in the plurality of candidate brands that matches the operating category identifier, determining all the plurality of candidate brands as the candidate brand set of the site; splicing the site feature data and the brand feature data of each candidate brand in the candidate brand set to generate spliced feature data; and inputting the spliced feature data into the recommendation model, and outputting the recommended brand list corresponding to the site by the recommendation model. In this way, by dynamically selecting accurate matching or full-quantity matching according to whether the site has an operating category identifier, the matching accuracy is taken into account while ensuring the candidate set coverage, thereby improving the rationality of the recommendation model input data and the quality of the recommendation result.

[0010] Optionally, in a further possible implementation manner of the first aspect, the calling the pre-trained risk prediction model to process the operating data of the plurality of stores, and outputting the risk state of each store by the risk prediction model comprises: for the operating data of each store, extracting operating index values of the store in a plurality of historical time windows; calculating a contract residual ratio and a unit area sales according to the operating index values; inputting the operating index values, the contract residual ratio and the unit area sales into the risk prediction model, and outputting the risk state of the store by the risk prediction model. In this way, by introducing the operating index values in the plurality of historical time windows, the contract residual ratio and the unit area sales, the risk prediction model is enhanced to capture the operating trend of the store, thereby improving the accuracy of the risk state determination.

[0011] Optionally, in a further possible implementation manner of the first aspect, the obtaining the operating data of the plurality of stores, the brand feature data of the plurality of candidate brands and the site feature data of the plurality of sites comprises: determining a current calculation mode according to a preset scheduling strategy, wherein the current calculation mode comprises a full-quantity calculation mode and an incremental calculation mode; in a case where the current calculation mode is the full-quantity calculation mode, obtaining the site feature data of all the sites and the brand feature data of the plurality of candidate brands; and in a case where the current calculation mode is the incremental calculation mode, in response to a data change event, determining an affected site corresponding to the data change event, and obtaining the site feature data of the affected site and the brand feature data of the plurality of candidate brands. In this way, by selecting to obtain all the site data or only the affected site data according to the scheduling strategy, unnecessary data obtaining and processing amount are reduced, thereby improving the data preparation efficiency and ensuring that the subsequent processing can respond to the latest changes in time.

[0012] Optionally, in a further possible implementation manner of the first aspect, the data change event includes at least one of a venue feature data change event, a brand feature data change event, and a risk state change event.

[0013] Optionally, in a further possible implementation manner of the first aspect, the intervention processing on the recommended brand list corresponding to the target venue according to the mapping relationship includes: in a case where the in-operation brand belongs to the preset non-recommended brand set, replacing the last brand in the recommended brand list with the in-operation brand, and marking the in-operation brand as a forced recommended brand in the updated recommended brand list. Thus, by replacing the last brand in the recommended brand list with the in-operation brand and adding a forced recommended mark in a case where the in-operation brand belongs to the preset non-recommended brand set, it is ensured that the risk warning intervention is not blocked by other business rules, and meanwhile, it is convenient to identify the intervention source in the front end, thereby improving controllability and traceability of the recommended result.

[0014] The second aspect of the embodiment of the present application provides a brand recommendation device, including: a data acquisition module configured to acquire operating data of a plurality of stores, brand feature data of a plurality of candidate brands, and venue feature data of a plurality of venues, wherein the brand feature data is used to represent operating attributes of the candidate brands, and the venue feature data is used to represent physical conditions and operating conditions of the venues.

[0015] a prediction model calling module configured to call a pre-trained risk prediction model to process the operating data of the plurality of stores, and the risk prediction model outputs a risk state of each store, wherein the risk state includes a low risk state and a high risk state.

[0016] a recommendation model calling module configured to call a pre-trained recommendation model to process the brand feature data of the plurality of candidate brands and the venue feature data of the plurality of venues, and the recommendation model outputs a recommended brand list corresponding to each venue.

[0017] a mapping establishing module configured to establish a mapping relationship between a target venue corresponding to a target store and an in-operation brand, wherein the target store is a store with a low risk state in the plurality of stores, the target venue is a venue where the target store is located, and the in-operation brand is a brand currently operated by the target store.

[0018] a brand recommendation module configured to perform intervention processing on the recommended brand list corresponding to the target venue according to the mapping relationship, to replace a last brand in the recommended brand list with the in-operation brand, to obtain an updated recommended brand list, and to output the updated recommended brand list.

[0019] The third aspect of the embodiments of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the brand recommendation method of the first aspect when executing the computer program.

[0020] The fourth aspect of the embodiments of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the brand recommendation method of the first aspect.

[0021] The fifth aspect of the embodiments of the present application provides a computer program product, which, when executed on an electronic device, causes the electronic device to execute the brand recommendation method of the first aspect.

[0022] It can be understood that the beneficial effects of the second aspect to the fifth aspect can be referred to the related description in the first aspect, which will not be repeated here. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0024] Figure 1 is a schematic diagram of the architecture of a brand recommendation system provided by the embodiments of the present application; Figure 2 is a flowchart of a brand recommendation method provided by the embodiments of the present application; Figure 3 is a structural schematic diagram of a brand recommendation device provided by the embodiments of the present application; Figure 4 is a structural schematic diagram of an electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0025] In the following description, specific details are set forth in order to provide a thorough understanding of the embodiments of the present application. However, persons skilled in the art should understand that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary details.

[0026] It should be understood that the word "comprise" or variations such as "comprises" or "comprising," when used in this specification and in the accompanying claims, specify the presence of stated features, integers, steps, or components but do not preclude the presence or addition of one or more other features, integers, steps, components, or groups thereof.

[0027] It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items, and that the term "at least one of' denotes one, or a plurality of, or any combination of the listed items.

[0028] As used in this specification and claims, the terms "if' and "when" can be interpreted to mean "upon determination" or "in response to a determination" or "in response to a detection" depending on the context.

[0029] In addition, the terms "first", "second", "third", etc. as used in the description of the specification and the appended claims are only used to distinguish different instances, and cannot be understood as indicating or implying relative importance.

[0030] Reference throughout this specification to "one embodiment" or "an embodiment" or "some embodiments" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. Thus, the appearances of the phrases "in one embodiment" or "in some embodiments" or "in other embodiments" or "in additional embodiments" or the like in various places throughout this specification are not necessarily all referring to the same embodiment, unless otherwise specified. The terms "including," "containing," "comprising," and variations thereof are meant to encompass the terms "consisting of" and "consisting essentially of" unless otherwise specified.

[0031] It should be understood that the size of the serial number of each step in the embodiments does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the application.

[0032] In the related art, the information on which the traditional brand recommendation is based focuses on the matching of the brand and the site itself, and the reflection capability of the store operation state change is limited. When the store operation state fluctuates, the original recommendation result may not be able to adapt to such change in time, and there is a deviation between the brand in the recommendation list and the actual operation demand, so the commercial personnel still need to rely on manual judgment for adjustment, affecting the timeliness and accuracy of the recommendation result.

[0033] Therefore, the embodiments of the present application provide a brand recommendation method, device and electronic equipment, wherein the method first acquires operation data of a plurality of stores, brand feature data of a plurality of candidate brands and site feature data of a plurality of sites, wherein the brand feature data is used to represent the operation attribute of the candidate brand, and the site feature data is used to represent the physical condition and operation condition of the site; then a pre-trained risk prediction model is called to process the operation data of the plurality of stores, and the risk prediction model outputs the risk state of each store, wherein the risk state includes a low risk state and a high risk state; then a pre-trained recommendation model is called to process the brand feature data of the plurality of candidate brands and the site feature data of the plurality of sites, and the recommendation model outputs a recommendation brand list corresponding to each site; further, a mapping relationship between the target site corresponding to the target store and the in-operation brand is established, wherein the target store is a store with a low risk state in the plurality of stores, the target site is the site where the target store is located, and the in-operation brand is the brand currently operated by the target store; finally, according to the mapping relationship, intervention processing is performed on the recommendation brand list corresponding to the target site to replace the last brand in the recommendation brand list with the in-operation brand, to obtain an updated recommendation brand list, and the updated recommendation brand list is output. Thus, by forcibly retaining the in-operation brand of the low-risk store in the recommendation list of the corresponding site, the recommendation result can automatically reflect the current risk state of the store, thereby improving the matching degree of the recommendation result and the actual operation demand, reducing manual adjustment, and improving the timeliness and accuracy of the commercial decision.

[0034] The application scenario of the brand recommendation method provided by the embodiments of the present application is exemplified below. Taking a shopping center as an example, in the process of commercial operation, the operator usually needs to generate a brand recommendation list for each stall to assist in commercial decision-making. In actual operation, the existing brand of some stalls is stable and has a low risk of withdrawing from the market, but the recommendation list calculated by the recommendation system according to the brand and site characteristics may not contain the in-operation brand, resulting in inconsistency between the recommendation result and the actual operation state of the stall. After adopting the solution of the present application, the store with a low risk state can be identified, and the in-operation brand corresponding to the store is forcibly retained in the recommendation list of the site where the store is located, so that the recommendation result takes into account the stability of the current operation of the stall, reduces the workload of manual comparison and adjustment of the commercial personnel, and improves the matching degree of the recommendation result and the actual operation demand.

[0035] To illustrate the technical solution of this application, specific embodiments are described below.

[0036] Reference Figure 1 The diagram illustrates the architecture of a brand recommendation system provided in an embodiment of this application. Figure 1 As shown, the brand recommendation system includes a withdrawal risk warning subsystem, a brand placement recommendation subsystem, a linkage processing module, and a front-end display module.

[0037] The store closure risk warning subsystem is used to call a pre-trained risk prediction model to process the acquired operating data from multiple stores and output the risk status of each store, including low-risk and high-risk statuses. Operating data refers to the data generated by the store during its operation.

[0038] The brand location recommendation subsystem calls a pre-trained recommendation model to process the brand feature data of multiple candidate brands and the site feature data of multiple venues, outputting a list of recommended brands for each venue. Brand feature data characterizes the operational attributes of candidate brands, while site feature data characterizes the physical and operational conditions of the venue.

[0039] The linkage processing module is used to identify target stores with low risk status from multiple stores based on risk status, establish a mapping relationship between the target store's target venue and the operating brands, and perform intervention processing on the recommended brand list based on this mapping relationship to obtain an updated recommended brand list.

[0040] The front-end display module is used to show the updated list of recommended brands and the risk status of stores.

[0041] exist Figure 1 Based on the system architecture shown, Figure 2 A flowchart illustrating a brand recommendation method provided in an embodiment of this application is shown. Figure 2 As shown, the brand recommendation method may include the following steps: Step 201: Obtain operating data of multiple stores, brand characteristic data of multiple candidate brands, and site characteristic data of multiple venues.

[0042] In the embodiments of the present application, the obtained operation data refers to data generated by the store in the operation process, and specifically can include store basic attribute information, contract state information, and historical time series data of multiple operation indexes. The store basic attribute information can include store identifier, store name, and belonging site identifier, etc. The contract state information can include contract start date, contract expiration date, contract term, etc. The historical time series data of operation indexes can include sales, sales quantity, unit area sales, rent-sales ratio, amount of arrears, performance guarantee, contract expiration date, etc. Through these operation data, the operation status and contract survival status of the store can be reflected, and data basis is provided for subsequent risk prediction.

[0043] In the embodiments of the present application, the brand feature data is used to represent the operation attribute of the candidate brand, and specifically can include brand headquarters information, operation scale index, brand heat index, cooperation mode information, operation category information, etc. The brand headquarters information can refer to the registered place or the location of the headquarters of the brand belonging enterprise. The operation scale index can refer to the number of stores and the number of employees of the brand. The brand heat index can refer to the search heat and attention of the brand on the preset platform. The cooperation mode information can refer to the cooperation mode between the brand and the commercial site, such as direct operation, franchise, etc. The operation category information can refer to the category of the goods or services operated by the brand. These feature data can depict the operation characteristics of the candidate brand from multiple dimensions, and are used for subsequent brand and site matching.

[0044] In the embodiments of the present application, the site feature data is used to represent the physical condition and operation condition of the site, and specifically can include site space information, geographic location information, operation category information, management personnel information, etc. The site space information can refer to the use area, shape, layer, etc. of the site. The geographic location information can refer to the floor, location number, adjacent passage, etc. of the site. The operation category information can refer to the current allowed operation category range or historical operation category of the site. The management personnel information can refer to the commercial recruitment personnel or operation personnel responsible for the site. These feature data are used to describe the basic situation and operation environment of the site, and provide basis for subsequent brand positioning recommendation.

[0045] In an embodiment, in the process of obtaining the above data, different data obtaining manners can be selected according to actual operation needs. Specifically, step 201 can specifically include: determining a current calculation mode according to a preset scheduling strategy, wherein the current calculation mode includes a full calculation mode and an incremental calculation mode; in a case where the current calculation mode is the full calculation mode, obtaining the site feature data of all sites and obtaining the brand feature data of the plurality of candidate brands; in a case where the current calculation mode is the incremental calculation mode, in response to a data change event, determining an affected site corresponding to the data change event, and obtaining the site feature data of the affected site and the brand feature data of the plurality of candidate brands. Further, the obtained site feature data of the affected site can be updated to the local cache data. In this way, by selecting to obtain all site data or only affected site data according to the scheduling strategy, unnecessary data obtaining and processing amount are reduced, thereby improving data preparation efficiency and ensuring that subsequent processing can respond to the latest changes in time.

[0046] It should be noted that the preset scheduling strategy is used to control the timing and range of data obtaining. The full calculation mode can be triggered according to a first preset period, for example, triggered once a day, at which time the site feature data of all sites and the brand feature data of the plurality of candidate brands are read from the data warehouse to ensure global consistency of data obtaining. The incremental calculation mode can be triggered according to a second preset period, for example, triggered once an hour, at which time instead of obtaining all data, the affected site is determined according to the data change event, and only the site feature data of the affected site is obtained, thereby reducing the amount of data obtaining and improving the timeliness of data processing. Wherein the first preset period is greater than the second preset period. In this way, while ensuring global consistency of data, re-obtaining of full data each time is avoided, and the consumption of computing resources is reduced.

[0047] Further, in the incremental calculation mode, the data change event is used to indicate which data has changed. Specifically, the data change event can include at least one of a site feature data change event, a brand feature data change event, and a risk state change event.

[0048] Wherein, the site feature data change event can mean that any field in the site feature data is added, modified or deleted; the brand feature data change event can mean that any field in the brand feature data is added, modified or deleted; the risk state change event can mean that the risk state output by the risk prediction model changes, for example, the risk state of a certain store changes from a low risk state to a high risk state, or from a high risk state to a low risk state. These data change events can be obtained from a data change log table, which is used to record various data changes occurring within a preset time window.

[0049] In a possible implementation, in the incremental calculation mode, the specific range of the affected site can be determined according to the type of the data change event. For a site feature data change event, the affected site is the site directly pointed to by the data change event, for example, when the use area or operation category information of a site changes, only the site is the affected site. For a brand feature data change event, the affected site includes all sites associated with the candidate brand whose information changes, for example, when the operation category information of a candidate brand changes, all related sites corresponding to the operation category information need to be determined as the affected site to avoid distortion of the recommendation result of multiple sites caused by the change of a single brand feature. For a risk state change event, the affected site is the site where the shop whose risk state changes is located, for example, when the risk state of a shop changes from a high-risk state to a low-risk state, the site where the shop is located is the affected site, and the site needs to be re-executed for subsequent recommendation model processing and intervention processing. By respectively determining the range of the affected site according to the type of the change event, the associated sites can be avoided from being missed while unnecessary reprocessing of all sites is avoided, the accuracy and processing efficiency of the incremental calculation are improved.

[0050] In step 202, a pre-trained risk prediction model is called to process the operation data of the plurality of shops, and the risk prediction model outputs the risk state of each shop.

[0051] The risk state includes a low-risk state and a high-risk state. The low-risk state indicates that the shop has a low possibility of withdrawing in the current stage, and the high-risk state indicates that the shop has a high possibility of withdrawing in the current stage.

[0052] In the embodiment of the present application, the risk prediction model is used to predict the risk probability of the shop withdrawing in the current stage according to the operation data of the shop, and to determine the risk state of the shop according to the risk probability. The risk prediction model can be a model trained by using a machine learning classification algorithm, and the input of the model is the feature extracted from the operation data, and the output of the model is the withdrawal risk probability of the shop. By calling the risk prediction model, the dispersed operation data can be converted into a structured risk state label, and a basis is provided for subsequent brand recommendation and intervention processing.

[0053] In an embodiment, step 202 can specifically include: for the operation data of each store, extracting operation indicator values of the store under a plurality of historical time windows; calculating the contract remaining ratio and the unit area sales according to the operation indicator values; inputting the operation indicator values, the contract remaining ratio and the unit area sales into a risk prediction model, and the risk prediction model outputting a risk state of the store. In this way, by introducing the operation indicator values of the plurality of historical time windows and the contract remaining ratio and the unit area sales, the risk prediction model's ability to capture store operation trends is enhanced, thereby improving the accuracy of risk state determination.

[0054] It should be noted that in the above implementation, the store is taken as the grouping unit, and the operation data of each store is extracted for features. Before extracting the operation indicator values, a reference record can be selected according to a time sequence, and the reference record is used to determine reference time points of the historical time windows. For example, the current statistical month is taken as the reference time, and 1 month, 2 months, 3 months, 6 months and 9 months are respectively traced back to obtain a plurality of historical time windows. For each historical time window, the operation indicator values corresponding to the time window are extracted, and the operation indicators can include sales, sales quantity, unit area sales, rent-sales ratio, amount of arrears, performance guarantee, contract expiration date, etc. Among them, the operation indicator values corresponding to each historical time window can be the indicator values at the end time node of the time window, or the average value or cumulative value of the indicators in the time window, which can be determined according to actual business needs. By extracting the operation indicator values under the plurality of historical time windows, the changing trends of the store operation in different time dimensions can be reflected, and more comprehensive input information is provided for the risk prediction model.

[0055] Among them, after obtaining the operation indicator values under the plurality of historical time windows, the contract remaining ratio and the unit area sales can be further calculated according to the operation indicator values. The contract remaining ratio and the unit area sales belong to derived features. Derived features refer to features obtained by operation or combination between indicators on the basis of original operation indicator values, which are used to reflect the operation of the store from different angles. The contract remaining ratio is used to represent the proportion of the contract remaining time length to the total contract time length, which can be calculated according to the contract expiration date and the signing date. For example, the contract remaining ratio is obtained by dividing the contract expiration date by the signing date. The unit area sales is used to represent the sales generated by each unit of the store, which can be calculated according to the sales and the use area. For example, the unit area sales is obtained by dividing the sales by the use area. By introducing the contract remaining ratio and the unit area sales, the contract survival pressure and the site operation efficiency of the store can be more directly reflected, which helps to improve the risk prediction model's ability to identify the risk of withdrawing the store.

[0056] After obtaining the operating indicator values, the contract residual ratios and the unit area sales of the plurality of historical time windows, the obtained feature data can be cleaned to remove sample records with missing historical features. For example, if the operating indicator values of a store in all historical time windows are empty, the sample record corresponding to the store does not participate in subsequent model prediction, so as to avoid interference of missing values on the prediction results.

[0057] Further, the operating indicator values, the contract residual ratios and the unit area sales are input into a pre-trained risk prediction model as input features. The risk prediction model outputs a store withdrawal risk probability according to the input features. According to a comparison result between the withdrawal risk probability and a preset threshold, a risk state of the store is determined. When the withdrawal risk probability is greater than the preset threshold, the risk state of the store is determined as a high-risk state; and when the withdrawal risk probability is not greater than the preset threshold, the risk state of the store is determined as a low-risk state. The preset threshold can be determined according to historical data statistical distribution or business experience, for example, can be set to 0.5, or can be adjusted according to actual operation requirements.

[0058] In some embodiments, for the store with the high-risk state, a withdrawal risk verification task can be automatically triggered and pushed to the marketing personnel responsible for the store, so that the marketing personnel can learn the operating risk of the store in time and take corresponding measures. For the store with the low-risk state, the risk state will be used for subsequent brand recommendation intervention processing, and the specific process will be described in subsequent steps.

[0059] In step 203, a pre-trained recommendation model is called to process the brand feature data of the plurality of candidate brands and the site feature data of the plurality of sites, and the recommendation model outputs a recommended brand list corresponding to each site.

[0060] In the embodiments of the present application, the recommendation model is used to predict the matching degree between each candidate brand and each site according to the brand feature data of the candidate brands and the site feature data of the sites, and generate a recommended brand list corresponding to each site according to the matching degree. The recommended brand list refers to a set of a plurality of candidate brands arranged from high to low according to the matching degree for a certain site, which is used to provide brand positioning reference for the marketing personnel.

[0061] In some embodiments, before the candidate brand set is constructed, a global non-recommended brand set can also be read from a brand restriction list data source, and the candidate brands in the global non-recommended brand set are excluded from the full brand library during the construction of the candidate brand set. The global non-recommended brand set maintained in the brand restriction list data source is the preset non-recommended brand set, and the brand restriction list data source is used to maintain a list of candidate brands that are not recommended under the current business rules. The global non-recommended brand set refers to the set of candidate brands recorded in the list. By excluding the candidate brands in the global non-recommended brand set, these candidate brands can be prevented from entering the subsequent matching and recommendation process, reducing invalid calculations and ensuring that the recommendation results meet the business constraints.

[0062] The brand restriction list can be maintained by the business development and operation personnel according to brand cooperation disputes, brand management abnormalities, or business strategies, and is updated regularly. The data in the brand restriction list data source is read before each execution of the recommendation process to ensure that the recommendation results reflect the latest business rules.

[0063] In some embodiments, step 203 can specifically include: for each site, if the site has an operating category identifier and there is a candidate brand in the plurality of candidate brands that matches the operating category identifier, determining the candidate brand that matches the operating category identifier as the candidate brand set of the site; if the site does not have an operating category identifier or there is no candidate brand in the plurality of candidate brands that matches the operating category identifier, determining all the plurality of candidate brands as the candidate brand set of the site; splicing the site feature data with the brand feature data of each candidate brand in the candidate brand set to generate spliced feature data; and inputting the spliced feature data into the recommendation model, and the recommendation model outputs a recommended brand list corresponding to the site. In this way, by dynamically selecting accurate matching or full matching according to the presence or absence of the operating category identifier of the site, the matching accuracy is taken into account while ensuring the coverage of the candidate set, thereby improving the rationality of the recommendation model input data and the quality of the recommendation results.

[0064] It should be noted that the operating category identifier is used to uniquely identify the product or service category allowed to be operated by the site, and can correspond to a specific subcategory level. The candidate brand set refers to the set of candidate brands participating in the matching calculation of a site. When constructing the candidate brand set, a category matching adaptive strategy is adopted.

[0065] Specifically, when the site has a business category identifier, and there are candidate brands in the plurality of candidate brands whose business category information in the brand feature data is consistent with the business category identifier, the candidate brands are determined as the candidate brand set of the site, that is, the exact matching mode is adopted, so that the candidate brands in the candidate brand set are consistent with the business category identifier of the site. When the site does not have a business category identifier, or although the site has a business category identifier, there are no candidate brands in the plurality of candidate brands whose business category information is consistent with the business category identifier, all candidate brands in the full brand library are determined as the candidate brand set of the site, that is, the full matching mode is adopted, avoiding the candidate brand set being empty due to incomplete category coverage, and ensuring that the recommendation process can be normally executed. Through the category matching adaptive strategy, when the site business category is clear and the brand library covers the corresponding category, exact matching is realized, and when the business category is missing or the brand library does not have the corresponding category, it is automatically downgraded to full matching, which balances the matching accuracy and the coverage of the candidate set.

[0066] Further, after determining the candidate brand set, the site feature data is spliced with the brand feature data of each candidate brand in the candidate brand set. The spliced feature data refers to the data obtained by combining the site feature data of the same site with the brand feature data of each candidate brand in the candidate brand set, which is used as the input of the recommendation model. Before splicing, the site feature data and the brand feature data can be subjected to feature preprocessing, which includes data type conversion and irrelevant feature elimination. Data type conversion refers to converting non-numeric data into numeric representation for processing by the recommendation model; irrelevant feature elimination refers to deleting fields irrelevant to brand and site matching from the feature data to reduce redundant information. Through feature preprocessing, the processing efficiency and prediction stability of the recommendation model can be improved.

[0067] Finally, the spliced feature data is input into the pre-trained recommendation model, and the recommendation model outputs the recommended brand list corresponding to the site. The recommendation model is a pre-trained machine learning classification model, which is used to output the matching degree between each candidate brand and the site according to the input feature data, and the matching degree can be represented as a recommendation score. The recommendation score is used to represent the adaptation degree of the candidate brand in the site, and the higher the recommendation score, the more suitable the candidate brand for the site. After the recommendation model outputs the recommendation score, the candidate brands in the candidate brand set can be sorted in descending order of the recommendation score, and a preset number of candidate brands with high ranking are selected to form the recommended brand list of the site.

[0068] In some embodiments, business constraint rules can also be applied to filter the recommendation results in the process of obtaining the recommended brand list. The business constraint rules can include a brand distribution balance constraint and a recommendation quantity constraint. The brand distribution balance constraint is used to limit the number of venues to which a same candidate brand is recommended, i.e., each candidate brand is referenced by at most a preset number of venues, so as to avoid excessive concentration of the same brand in commercial venues. The recommendation quantity constraint is used to limit the number of recommended brands reserved by each venue, i.e., each venue reserves at most a preset number of recommended brands.

[0069] In a possible implementation, the candidate brands in the candidate brand set can be traversed in the order of recommendation scores from high to low in the venue as a traversal unit, and the candidate brands are added to the recommended brand list of the venue under the premise that the brand distribution balance constraint and the recommendation quantity constraint are met, until the number of brands in the recommended brand list reaches the preset number corresponding to the recommendation quantity constraint. By applying the business constraint rules, the recommendation results can meet the overall balance requirement of business operation while meeting the model prediction.

[0070] The preset number in each constraint described above can be determined according to the operation scale of the commercial venue, the brand layout strategy and other factors. For example, for the brand distribution balance constraint, the preset number can be set to 3, indicating that a same candidate brand is referenced by at most 3 venues. For the recommendation quantity constraint, the preset number can be set to 5, indicating that each venue reserves at most 5 recommended brands, i.e., a Top-5 recommended brand list is generated. It should be noted that the above values are only examples, and can be flexibly configured in actual application.

[0071] In the embodiments of the present application, the candidate brand set construction, feature splicing, model prediction and business constraint filtering are performed in the venue as a traversal unit. When the full calculation mode is adopted, the complete process described above is performed on all venues. When the incremental calculation mode is adopted, the process described above is only performed on the affected venues, and the specific data acquisition manner has been described in step 201, which will not be repeated here. Through the above processing, the recommendation model outputs the recommended brand list corresponding to each venue, which is used for subsequent intervention processing.

[0072] In an embodiment, the training samples of the recommendation model can include brand and site combinations that have successfully settled and operated for more than a preset length of time historically, and brand and site combinations that have settled for a short period of time historically and then withdrawn or performed poorly. By training a machine learning classification model using these historical samples, a recommendation model that can predict the matching degree of brand and site is obtained. The training features of the recommendation model can include all or part of the fields in the brand feature data and the site feature data, and the training labels can include whether the combination belongs to a successful match. By using the above training method, the recommendation model can learn the matching rules between brand features and site features, providing a basis for subsequent recommendations.

[0073] Step 204, a mapping relationship between the target site corresponding to the target store and the in-operation brand is established.

[0074] Among them, the target store is a store with a low risk state in the plurality of stores, the target site is the site where the target store is located, and the in-operation brand is the brand currently operated by the target store.

[0075] In the embodiment of the present application, the mapping relationship is used to associate the target site with the brand currently operating in the target site, and to provide a basis for subsequent recommendation brand list intervention processing. By establishing the mapping relationship, the output result of the risk prediction model can be connected with the processing process of the recommendation brand list, so that the in-operation brand corresponding to the store with a low risk state can be retained in the recommendation brand list.

[0076] In an embodiment, step 204 can specifically include: selecting stores with a low risk state from the plurality of stores to obtain a target store set; batch acquiring the site identifier and the brand identifier of each store in the target store set by calling a preset external interface; and establishing a mapping relationship between each target site and the corresponding in-operation brand according to the site identifier and the brand identifier. In this way, by calling the external interface to batch acquire the site identifier and the brand identifier and establishing the mapping relationship, the manual query link is reduced, thereby improving the efficiency and accuracy of mapping relationship construction, and enabling the risk early warning result to timely drive subsequent recommendation intervention.

[0077] It should be noted that the target store set refers to a set of all stores with a low risk state selected from the risk states output by the risk prediction model. In the selection process, the risk state of each store output by the risk prediction model can be traversed, and the stores with a low risk state are added to the target store set. By selecting low-risk stores, rather than establishing a mapping relationship for all stores one by one, subsequent recommendation intervention can be performed only on stores with stable operation and low withdrawal risk, reducing unnecessary mapping relationship construction overhead, and making the recommendation intervention more targeted.

[0078] The site identifier is used to uniquely identify a site, and the brand identifier is used to uniquely identify the specific brand identity of a candidate brand in the current operating state. After determining the target store set, the site identifier and the brand identifier currently operated by each store in the target store set are batch acquired by calling a preset external interface. The external interface refers to a data query interface connected with the operation management system or the master data management system of the commercial site, which is used to query the site identifier and the brand identifier currently operated by the store according to the store identifier. By batch acquisition through calling the external interface instead of manual query, the efficiency and accuracy of the mapping relationship construction can be improved.

[0079] It should be noted that the brand identifier is used to uniquely identify a candidate brand. The in-operation brand refers to the brand currently actually operated by the target store, which belongs to one of the candidate brands. However, in the brand recommendation process, the in-operation brand may appear in the recommended brand list or may not be output by the recommendation model. The mapping relationship is associated with the brand identifier corresponding to the in-operation brand, so that the corresponding position of the in-operation brand in the recommended brand list can be accurately found for subsequent intervention processing.

[0080] Further, after establishing the mapping relationship according to the site identifier and the brand identifier, the mapping relationship between each target site and the corresponding in-operation brand is obtained. The mapping relationship can be stored in the form of key-value pairs, in which the key is the site identifier of the target site, and the value is the brand identifier of the in-operation brand corresponding to the target site. Subsequently, when the recommended brand list is intervened, the corresponding brand identifier can be found in the mapping relationship according to the site identifier of the target site, so as to determine the in-operation brand to be retained. Through the above method, after the risk prediction model outputs the risk state, the mapping relationship can be automatically and batch constructed, providing a data basis for the closed loop of risk prediction and recommendation intervention.

[0081] In step 205, according to the mapping relationship, intervention processing is performed on the recommended brand list corresponding to the target site to replace the last brand in the recommended brand list with the in-operation brand, to obtain an updated recommended brand list, and output the updated recommended brand list.

[0082] In the embodiments of the present application, the intervention processing refers to modifying the recommended brand list output by the recommendation model according to the mapping relationship, so that the in-operation brand corresponding to the target store with a low risk state can be retained in the recommended brand list of the target site. Through the intervention processing, the recommended brand list no longer depends only on the matching score between the brand and the site, but is combined with the actual operating risk state of the store, so that the recommended result is more in line with the current operation of the commercial site.

[0083] In an embodiment, if the in-operation brand is already included in the recommended brand list, the operation of replacing the last brand in the recommended brand list with the in-operation brand can be skipped; if the in-operation brand is not included in the recommended brand list, the last brand in the recommended brand list is replaced with the in-operation brand.

[0084] It should be noted that when the recommended brand list corresponding to a target site is intervened according to the mapping relationship, it is first determined whether the in-operation brand corresponding to the target site already exists in the recommended brand list. If the in-operation brand is already included in the recommended brand list, it means that the recommended model has already listed the in-operation brand as a recommended object, and at this time, no replacement is needed, and the replacement operation is directly skipped, and the original recommended brand list is retained. If the in-operation brand is not included in the recommended brand list, the last brand in the recommended brand list is replaced with the in-operation brand. By first determining and then replacing, unnecessary repeated replacement of the recommended brand list that already contains the in-operation brand can be avoided, the execution efficiency of the intervention processing is improved, and the stability of the recommended brand list is maintained.

[0085] In an embodiment, if the in-operation brand belongs to a preset non-recommended brand set, the last brand in the recommended brand list is replaced with the in-operation brand, and the in-operation brand is marked as a forced recommended brand in the updated recommended brand list. In this way, by still performing replacement and adding a forced recommendation mark when the in-operation brand belongs to the preset non-recommended brand set, it is ensured that the risk warning intervention is not blocked by other business rules, and at the same time, it is convenient for the front end to identify the intervention source, thereby improving the controllability and traceability of the recommended result.

[0086] It should be noted that the non-recommended brand set refers to a candidate brand set that is not recommended for recommendation according to business rules, for example, a brand set that is listed in a restriction list due to brand operation abnormalities, cooperation disputes, etc. When the mapping relationship is established, if it is determined that the in-operation brand corresponding to the target site belongs to the preset non-recommended brand set, the in-operation brand is still replaced to the last position in the recommended brand list, and the in-operation brand is marked as a forced recommended brand in the updated recommended brand list. The forced recommended brand is used to indicate that the in-operation brand is forced to be retained in the recommended brand list based on the risk warning intervention, rather than being naturally recommended by the recommendation model according to the matching score. By adding this mark at the data level, the front-end display module can identify the intervention source accordingly, and display the brand as an in-operation continuation promotion identifier, which facilitates the business personnel to distinguish between normal recommended brands and in-operation brands retained due to risk warning intervention.

[0087] In the intervention processing, the replacement operation keeps the number of brands in the recommended brand list unchanged, that is, only the last brand in the recommended brand list is replaced by an in-operation brand, and the number of brands in the recommended brand list is not increased or decreased. When intervention processing is performed on multiple target sites, the replacement can be performed on a site-by-site basis according to the mapping relationship until all target sites that hit the mapping relationship complete the intervention processing. Through the above intervention processing, an updated recommended brand list is finally obtained, and the updated recommended brand list is output.

[0088] In a possible implementation, after obtaining the updated recommended brand list, the updated recommended brand list can be written into a recommended result table for reading and display by a front-end display module. The recommended result table is used to store the recommended brand list corresponding to each site and related attributes thereof. According to different current calculation modes, the updating manner of the recommended result table can be different.

[0089] In a possible implementation, in a case where the current calculation mode is the full calculation mode, the historical recommended brand list is deleted, and the updated recommended brand list is written into the recommended result table in batches. The historical recommended brand list refers to the recommended brand list stored in the recommended result table after the last full calculation or incremental calculation. Through the manner of deleting historical data first and then writing in batches, it can be ensured that the data in the recommended result table is completely consistent with the result of the full calculation, and data inconsistency caused by residual historical data is avoided.

[0090] In a possible implementation, in a case where the current calculation mode is the incremental calculation mode, the scores of brands in the historical recommended brand list of the affected site are set to zero, the historical prediction scores are updated according to the recommendation score increment obtained by the current calculation, and the updated recommended brand list is written row by row. Through the manner of setting the historical recommendation scores of the affected site to zero first and then updating row by row, it can be avoided that there are new and old records of the same site in the incremental updating process, and the data consistency of the recommended result table in the updating process is ensured. In the incremental calculation mode, only the records of the affected site are updated, and full data does not need to be deleted and rewritten, so the data writing amount can be significantly reduced, and the updating efficiency is improved.

[0091] In some embodiments, after the updating of the recommended result table is completed, the front-end display module can read the updated recommended brand list from the recommended result table and display it. The content displayed by the front-end display module can include the recommended brand list corresponding to each site, the ranking of the recommendation scores of brands in the recommended brand list, brand distribution statistical information, and the like. For the in-operation brand marked as a forced recommended brand in the intervention processing, the front-end display module can display it as an “in-operation continuation recommendation” identifier when displaying, so that the marketing personnel can intuitively distinguish the normal recommended brand from the in-operation brand retained due to the risk early warning intervention.

[0092] In some embodiments, the front-end display module can also display risk warning information, including the risk state of each store, the risk probability of withdrawal, the number of days to contract expiration, and the like output from the risk prediction model. The marketing personnel can view the risk state of each store and the corresponding recommended brand list through the front-end display module, and perform data filtering, sorting, exporting, detail viewing, and the like as needed to assist in marketing decision-making.

[0093] In some embodiments, the calling of the risk prediction model can be performed according to a preset monthly scheduling period, that is, the withdrawal risk warning processing flow is triggered once a month, the operating data of each store in the month is analyzed, and the risk state is updated. After updating the risk state, the stores with changed risk state are written into the data change log table as risk state change events, thereby triggering the data change events in the incremental calculation mode, so that the recommended brand list can respond to the change of the risk state in a timely manner, thereby maintaining the linkage timeliness between the risk warning and the brand recommendation.

[0094] The brand recommendation method disclosed in the above embodiments of the present application first acquires the operating data of a plurality of stores, brand feature data of a plurality of candidate brands, and site feature data of a plurality of sites, wherein the brand feature data is used to represent the operating attributes of the candidate brands, and the site feature data is used to represent the physical conditions and operating conditions of the sites; then a pre-trained risk prediction model is called to process the operating data of the plurality of stores, and the risk prediction model outputs the risk state of each store, wherein the risk state includes a low-risk state and a high-risk state; then a pre-trained recommendation model is called to process the brand feature data of the plurality of candidate brands and the site feature data of the plurality of sites, and the recommendation model outputs the recommended brand list corresponding to each site; further, a mapping relationship between the target site corresponding to the target store and the in-operation brand is established, wherein the target store is a store with a low-risk state in the plurality of stores, the target site is the site where the target store is located, and the in-operation brand is the brand currently operated by the target store; finally, according to the mapping relationship, intervention processing is performed on the recommended brand list corresponding to the target site to replace the last brand in the recommended brand list with the in-operation brand, to obtain an updated recommended brand list, and the updated recommended brand list is output. Thus, by forcibly retaining the in-operation brand of the low-risk store in the recommended list of the corresponding site, the recommended result can automatically reflect the current risk state of the store, thereby improving the matching degree of the recommended result and the actual operating demand, reducing manual adjustment, and improving the timeliness and accuracy of marketing decision-making.

[0095] Referring to Figure 3 , a structural schematic diagram of a brand recommendation device provided by an embodiment of the present application is shown, and only parts related to the embodiments of the present application are shown for ease of illustration.

[0096] The brand recommendation device can specifically include the following modules: The data acquisition module 301 is configured to acquire operating data of a plurality of stores, brand feature data of a plurality of candidate brands, and site feature data of a plurality of sites, wherein the brand feature data is used to represent the operating attributes of the candidate brands, and the site feature data is used to represent the physical conditions and operating conditions of the sites.

[0097] The prediction model calling module 302 is configured to call a pre-trained risk prediction model to process the operating data of the plurality of stores, and the risk prediction model outputs a risk state of each store, wherein the risk state includes a low-risk state and a high-risk state.

[0098] The recommendation model calling module 303 is configured to call a pre-trained recommendation model to process the brand feature data of the plurality of candidate brands and the site feature data of the plurality of sites, and the recommendation model outputs a recommended brand list corresponding to each site.

[0099] The mapping establishing module 304 is configured to establish a mapping relationship between a target site corresponding to a target store and an in-operation brand, wherein the target store is a store with a low-risk state among the plurality of stores, the target site is a site where the target store is located, and the in-operation brand is a brand currently operated by the target store.

[0100] The brand recommendation module 305 is configured to perform intervention processing on the recommended brand list corresponding to the target site according to the mapping relationship, to replace the last brand in the recommended brand list with the in-operation brand, to obtain an updated recommended brand list, and to output the updated recommended brand list.

[0101] The brand recommendation device disclosed in the above embodiments of the present application first acquires the operating data of a plurality of stores, brand feature data of a plurality of candidate brands, and site feature data of a plurality of sites, wherein the brand feature data is used to represent the operating attributes of the candidate brands, and the site feature data is used to represent the physical conditions and operating conditions of the sites; then a pre-trained risk prediction model is called to process the operating data of the plurality of stores, and the risk prediction model outputs the risk state of each store, wherein the risk state includes a low-risk state and a high-risk state; then a pre-trained recommendation model is called to process the brand feature data of the plurality of candidate brands and the site feature data of the plurality of sites, and the recommendation model outputs a recommended brand list corresponding to each site; further, a mapping relationship between a target site corresponding to a target store and an in-operation brand is established, wherein the target store is a store with a low-risk state in the plurality of stores, the target site is a site where the target store is located, and the in-operation brand is a brand currently operated by the target store; finally, according to the mapping relationship, intervention processing is performed on the recommended brand list corresponding to the target site to replace the last brand in the recommended brand list with the in-operation brand, to obtain an updated recommended brand list, and the updated recommended brand list is output. In this way, by forcibly retaining the in-operation brand of the low-risk store in the recommended list of the corresponding site, the recommendation result can automatically reflect the current risk state of the store, thereby improving the matching degree of the recommendation result and the actual operating demand, reducing manual adjustment, and improving the timeliness and accuracy of the marketing decision.

[0102] Further, in a possible implementation manner of the embodiment of the present application, the brand recommendation module 305 can specifically include the following units: The first processing unit is configured to, in the case that the in-operation brand is already included in the recommended brand list, skip the operation of replacing the last brand in the recommended brand list with the in-operation brand.

[0103] The second processing unit is configured to, in the case that the in-operation brand is not included in the recommended brand list, replace the last brand in the recommended brand list with the in-operation brand.

[0104] In this way, by skipping the replacement when the recommended list already contains the in-operation brand, unnecessary repeated intervention on the existing result is avoided, thereby improving the execution efficiency of the intervention processing and ensuring the stability of the recommended list.

[0105] Further, in another possible implementation manner of the embodiment of the present application, the mapping establishment module 304 can specifically include the following units: The third processing unit is configured to filter the stores with a low-risk state from the plurality of stores to obtain a target store set.

[0106] The first obtaining unit is configured to obtain, in batches, the site identifier and the brand identifier currently operated by each store in the target store set by invoking a preset external interface.

[0107] The fourth processing unit is configured to establish a mapping relationship between each target site and the corresponding brand under operation according to the site identifier and the brand identifier.

[0108] Therefore, by invoking the external interface to obtain the site identifier and the brand identifier in batches and establishing the mapping relationship, the manual query link is reduced, the efficiency and accuracy of the mapping relationship construction are improved, and the risk early warning result can timely drive subsequent recommendation intervention.

[0109] Further, in another possible implementation manner of the embodiment of the application, the recommendation model calling module 303 can specifically include the following units: The first determining unit is configured to, for each site, in a case where the site has an operating category identifier and there is a candidate brand matching the operating category identifier in the plurality of candidate brands, determine the candidate brand matching the operating category identifier as the candidate brand set of the site.

[0110] The second determining unit is configured to, in a case where the site does not have an operating category identifier or there is no candidate brand matching the operating category identifier in the plurality of candidate brands, determine all the plurality of candidate brands as the candidate brand set of the site.

[0111] The first generating unit is configured to splice the site feature data and the brand feature data of each candidate brand in the candidate brand set to generate spliced feature data.

[0112] The fifth processing unit is configured to input the spliced feature data into the recommendation model, and the recommendation model outputs a recommended brand list corresponding to the site.

[0113] Therefore, by dynamically selecting accurate matching or full-quantity matching according to whether the site has an operating category identifier, the matching accuracy is taken into account while ensuring the candidate set coverage, so that the rationality of the input data of the recommendation model and the quality of the recommendation result are improved.

[0114] Further, in another possible implementation manner of the embodiment of the application, the prediction model calling module 302 can specifically include the following units: The sixth processing unit is configured to, for the operating data of each store, extract operating index values of the store in a plurality of historical time windows.

[0115] The first calculating unit is configured to calculate the contract residual ratio and the unit area sales according to the operating index values.

[0116] The seventh processing unit is configured to input the business index value, the contract residual ratio and the unit area sales into a risk prediction model, and the risk prediction model outputs a risk state of the store.

[0117] In this way, by introducing the business index value and the contract residual ratio and the unit area sales of multiple historical time windows, the risk prediction model can enhance the ability to capture the business trend of the store, thereby improving the accuracy of the risk state determination.

[0118] Further, in another possible implementation of the embodiment of the application, the data acquisition module 301 can specifically include the following units: The third determination unit is configured to determine a current calculation mode according to a preset scheduling strategy, wherein the current calculation mode includes a full calculation mode and an incremental calculation mode.

[0119] The eighth processing unit is configured to, in a case where the current calculation mode is the full calculation mode, acquire the site feature data of all sites and acquire the brand feature data of the plurality of candidate brands.

[0120] The ninth processing unit is configured to, in a case where the current calculation mode is the incremental calculation mode, in response to a data change event, determine an affected site corresponding to the data change event, acquire the site feature data of the affected site, and acquire the brand feature data of the plurality of candidate brands.

[0121] In this way, by selecting to acquire all site data or only affected site data according to the scheduling strategy, unnecessary data acquisition and processing amount are reduced, thereby improving the data preparation efficiency and ensuring that the subsequent processing can respond to the latest changes in time.

[0122] Further, in another possible implementation of the embodiment of the application, the data change event includes at least one of a site feature data change event, a brand feature data change event and a risk state change event.

[0123] Further, in another possible implementation of the embodiment of the application, the brand recommendation module 305 can specifically include the following units: The tenth processing unit is configured to, in a case where the in-operation brand belongs to a preset non-recommended brand set, replace the brand at the end of the recommended brand list with the in-operation brand, and mark the in-operation brand as a forced recommended brand in the updated recommended brand list.

[0124] In this way, by still performing replacement and adding a forced recommendation mark when the in-operation brand belongs to the preset non-recommended brand set, it is ensured that the risk warning intervention is not blocked by other business rules, and at the same time, it is convenient for the front end to identify the source of the intervention, thereby improving the controllability and traceability of the recommendation result.

[0125] The brand recommendation device provided by the embodiments of the present application can be applied in the method embodiments described above, and details can be referred to the description of the method embodiments, which will not be repeated here.

[0126] Figure 4 is a structural schematic diagram of an electronic device provided by the embodiments of the present application. As shown in the figure, Figure 4 The electronic device 400 of the embodiments includes at least one processor 410 (only one processor is shown in the figure), a memory 420, and a computer program 421 stored in the memory 420 and executable on the at least one processor 410, wherein the processor 410 implements the steps in the brand recommendation method embodiments described above when executing the computer program 421. Figure 4 The electronic device 400 can be a desktop computer, a notebook computer, a palm computer, a cloud server, and other computing devices. The electronic device can include, but is not limited to, the processor 410 and the memory 420. Those skilled in the art can understand that

[0127] The electronic device 400 is only an example and does not constitute a limitation on the electronic device 400, and can include more or fewer components than those shown in the figure, or combine certain components, or different components, for example, can also include an input / output device, a network access device, etc. Figure 4 The processor 410 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic components, discrete hardware components, etc. The general-purpose processor can be a microprocessor or can also be any conventional processor.

[0128]

[0129] ​The memory 420 can be an internal memory unit of the electronic device 400, such as a hard disk or a memory of the electronic device 400 in some embodiments. The memory 420 can also be an external memory device of the electronic device 400, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the electronic device 400 in some other embodiments. Further, the memory 420 can include both an internal memory unit and an external memory device of the electronic device 400. The memory 420 is used to store an operating system, an application program, a Boot Loader, data, and other programs, such as program codes of the computer program, etc. The memory 420 can also be used to temporarily store data that has been output or is to be output.

[0130] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is taken as an example for description, and in actual application, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit, and the integrated unit can be realized in the form of hardware or in the form of software functional unit. In addition, the specific names of each functional unit and module are only for the convenience of mutual distinction, and do not limit the protection scope of the present application. The specific working process of the units and modules in the system can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0131] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0132] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or in a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0133] In the embodiments of the present application, it should be understood that the disclosed apparatus / equipment and method can be implemented in other manners. For example, the described apparatus / equipment embodiments are merely schematic. For example, the division of the modules or units is merely logical function division. There can be another division manner for the actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0134] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0135] In addition, each functional unit in the various embodiments of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0136] The integrated module / unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the above-described embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. The computer program, when executed by a processor, can implement the steps of each method embodiment. The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the computer readable medium can include appropriate contents according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0137] The application implements all or part of the processes in the above-mentioned embodiment methods, and can also be completed by a computer program product. When the computer program product runs on an electronic device, the electronic device is caused to perform the steps in each of the above-mentioned method embodiments.

[0138] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A brand recommendation method, characterized in that, include: The system acquires operational data from multiple stores, brand characteristic data from multiple candidate brands, and site characteristic data from multiple locations. The brand characteristic data is used to characterize the operational attributes of the candidate brands, and the site characteristic data is used to characterize the physical and operational conditions of the locations. A pre-trained risk prediction model is invoked to process the operating data of the multiple stores. The risk prediction model outputs the risk status of each store, wherein the risk status includes a low-risk status and a high-risk status. A pre-trained recommendation model is invoked to process the brand feature data of the multiple candidate brands and the site feature data of the multiple sites. The recommendation model outputs a list of recommended brands for each site. Establish a mapping relationship between the target store and the target venue and the operating brand, wherein the target store is the store with a low risk status among the multiple stores, the target venue is the venue where the target store is located, and the operating brand is the brand currently operated by the target store; Based on the mapping relationship, an intervention process is performed on the recommended brand list corresponding to the target venue to replace the last brand in the recommended brand list with the operating brand, thereby obtaining an updated recommended brand list, and the updated recommended brand list is output.

2. The method according to claim 1, characterized in that, The step of performing intervention processing on the recommended brand list corresponding to the target venue according to the mapping relationship, so as to replace the last brand in the recommended brand list with the operating brand, includes: If the operating brand is already included in the recommended brand list, skip the operation of replacing the last brand in the recommended brand list with the operating brand; If the operating brand is not included in the recommended brand list, the last brand in the recommended brand list will be replaced with the operating brand.

3. The method according to claim 1, characterized in that, The process of establishing a mapping relationship between the target store's corresponding target location and the operating brands includes: From the multiple stores, stores with a low risk status are selected to obtain the target store set; By calling a preset external interface, the site identifier and the currently operating brand identifier of each store in the target store set can be obtained in batches. Based on the site identifier and the brand identifier, establish a mapping relationship between each target site and the corresponding operating brand.

4. The method according to claim 1, characterized in that, The process involves calling a pre-trained recommendation model to process the brand feature data of the multiple candidate brands and the site feature data of the multiple sites. The recommendation model then outputs a list of recommended brands for each site, including: For each of the aforementioned venues, if the venue has a business category identifier and there is a candidate brand among the plurality of candidate brands that matches the business category identifier, the candidate brand that matches the business category identifier is determined as the candidate brand set for the venue. If the venue does not have a business category identifier, or if there is no candidate brand among the multiple candidate brands that matches the business category identifier, all of the multiple candidate brands will be determined as the candidate brand set for the venue. The site feature data is spliced ​​with the brand feature data of each candidate brand in the candidate brand set to generate spliced ​​feature data. The stitched feature data is input into the recommendation model, and the recommendation model outputs a list of recommended brands corresponding to the venue.

5. The method according to claim 1, characterized in that, The process involves calling a pre-trained risk prediction model to process the operating data of the multiple stores. The risk prediction model outputs the risk status of each store, including: For each store's operating data, extract the store's operating indicator values ​​across multiple historical time windows; Calculate the contract remaining ratio and sales per unit area based on the aforementioned operating indicator values; The operating indicator values, the remaining contract ratio, and the sales revenue per unit area are input into the risk prediction model, and the risk prediction model outputs the risk status of the store.

6. The method according to claim 1, characterized in that, The acquisition of operational data from multiple stores, brand characteristic data from multiple candidate brands, and site characteristic data from multiple venues includes: According to the preset scheduling strategy, the current computing mode is determined, wherein the current computing mode includes full computing mode and incremental computing mode; When the current calculation mode is the full calculation mode, obtain the site feature data of all the sites, and obtain the brand feature data of the multiple candidate brands; When the current calculation mode is the incremental calculation mode, in response to a data change event, the affected site corresponding to the data change event is determined, and the site feature data of the affected site and the brand feature data of the multiple candidate brands are obtained.

7. The method according to claim 6, characterized in that, The data change event includes at least one of the following: site characteristic data change event, brand characteristic data change event, and risk status change event.

8. The method according to claim 1, characterized in that, The step of performing intervention processing on the recommended brand list corresponding to the target venue according to the mapping relationship, so as to replace the last brand in the recommended brand list with the operating brand, to obtain an updated recommended brand list, includes: If the operating brand belongs to a preset set of unrecommended brands, the last brand in the recommended brand list is replaced with the operating brand, and the operating brand is marked as a mandatory recommended brand in the updated recommended brand list.

9. A brand recommendation device, characterized in that, include: The data acquisition module is used to acquire operating data of multiple stores, brand characteristic data of multiple candidate brands, and site characteristic data of multiple venues. The brand characteristic data is used to characterize the operating attributes of the candidate brands, and the site characteristic data is used to characterize the physical and operating conditions of the venues. The prediction model invocation module is used to invoke a pre-trained risk prediction model to process the operating data of the multiple stores. The risk prediction model outputs the risk status of each store, wherein the risk status includes a low-risk status and a high-risk status. The recommendation model invocation module is used to invoke a pre-trained recommendation model to process the brand feature data of the multiple candidate brands and the site feature data of the multiple sites. The recommendation model outputs a list of recommended brands for each site. The mapping establishment module is used to establish a mapping relationship between the target store and the target venue and the operating brand. The target store is the store with a low risk status among the multiple stores, the target venue is the venue where the target store is located, and the operating brand is the brand currently operated by the target store. The brand recommendation module is used to perform intervention processing on the recommended brand list corresponding to the target venue according to the mapping relationship, so as to replace the last brand in the recommended brand list with the operating brand, obtain an updated recommended brand list, and output the updated recommended brand list.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 8.

Images