Method and device for determining shelf goods

Abstract

The present invention discloses a method and device for determining goods to be put on the shelves, the method comprising: determining a plurality of candidate goods to be put on the shelves; calculating the store characteristic parameters of the store corresponding to each of the candidate goods to be put on the shelves; inputting the plurality of candidate goods to be put on the shelves and the corresponding store characteristic parameters into the prediction network model corresponding to each of the candidate goods to be put on the shelves, so as to determine the target candidate goods to be put on the shelves; the prediction network model is obtained by training a training set including a plurality of stores with known sales effects and corresponding store characteristic parameters of the corresponding candidate goods to be put on the shelves. It can be seen that the present invention can be combined with artificial intelligence technology to assist in selecting goods to be put on the shelves. On the one hand, it can fully combine the store characteristic parameters of the store to consider the selection of goods to be put on the shelves, and on the other hand, it achieves an efficient and accurate effect of determining the goods to be put on the shelves.

Description

Technical Field

[0001] The present invention relates to the technical field of recommendation algorithms, and in particular to a method and device for determining shelf goods. Background Art

[0002] When commercial stores select new products to put on the shelves, the existing solutions generally use two methods to confirm the product selection strategy. One is to select suitable products based on the characteristics of the scene around the convenience store, such as daily chemical products when there are many residential communities around, and office supplies when there are many office buildings around. The second is to conduct store traffic surveys, judge consumer preferences by distributing questionnaires, collecting and analyzing data, and then select appropriate products to put on the shelves. But obviously, the method of manual inspection and product selection based on experience is very subjective. First of all, manual considerations cannot be comprehensive, and cannot fully judge the strong correlation between store traffic and surrounding scenes. There are inevitably limitations. Secondly, the manual decision-making method is inefficient. The method of conducting store traffic surveys requires the distribution of questionnaires, data collection and analysis, which requires a large amount of manual investment and is difficult to quickly replicate and promote when there are a large number of stores. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a method and device for determining goods to be put on the shelves, which can combine artificial intelligence technology to assist in selecting goods to be put on the shelves. On the one hand, it can fully combine the store characteristic parameters of the store to consider the selection of goods for shelving, and on the other hand, it can achieve efficient and accurate goods shelf determination effect.

[0004] In order to solve the above technical problems, the first aspect of the present invention discloses a method for determining shelf goods, the method comprising:

[0005] Identify multiple candidate products for listing;

[0006] Calculate the store characteristic parameters of the store corresponding to each candidate listing product;

[0007] The multiple candidate shelves and the corresponding store characteristic parameters are respectively input into the prediction network model corresponding to each of the candidate shelves to determine the target candidate shelves; the prediction network model is obtained by training with a training set including multiple stores with known sales effects and corresponding store characteristic parameters of the corresponding candidate shelves.

[0008] As an optional implementation, in the first aspect of the present invention, the store characteristic parameters include at least one of store surrounding traffic characteristics, store surrounding facility characteristics, and store surrounding competitive product characteristics.

[0009] As an optional implementation, in the first aspect of the present invention, the calculating of the store characteristic parameters of each candidate listing product includes:

[0010] Determine the surrounding traffic information of the listed store;

[0011] Determine the surrounding crowd flow portrait corresponding to the candidate listing product according to the surrounding crowd flow information and a preset crowd portrait matching database;

[0012] Determine the surrounding crowd flow portrait as the surrounding crowd flow feature of the store of the candidate product;

[0013] and / or,

[0014] Determine the surrounding traffic information of the listed store;

[0015] Determine a number of target facilities and a number of competing stores corresponding to the listed store;

[0016] Determine the surrounding traffic information of the target facility and the competing store;

[0017] Calculate the crowd flow overlap between the surrounding crowd flow information of the listed store and the surrounding crowd flow information of any of the target facilities or the competing stores;

[0018] The overlap of all the crowds is determined as the crowd flow characteristics around the store of the candidate listing product.

[0019] As an optional implementation, in the first aspect of the present invention, the calculating of the store characteristic parameters of each candidate listing product includes:

[0020] Determine a plurality of target facilities around the location of the listed store;

[0021] Determine a distance parameter between the location of the listed store and any of the target facilities;

[0022] The distance parameters corresponding to all the target facilities are determined as the store surrounding facility features of the candidate shelf products.

[0023] As an optional implementation, in the first aspect of the present invention, the calculating of the store characteristic parameters of each candidate listing product includes:

[0024] Determine multiple competing stores around the location of the listed store;

[0025] Determine a distance parameter between the location of the listing store and any of the competing stores;

[0026] The distance parameters corresponding to all the competing product stores are determined as the competing product features around the stores of the candidate listed products.

[0027] As an optional implementation, in the first aspect of the present invention, the surrounding human flow information is device information of all users within the coverage area of ​​the corresponding human flow.

[0028] As an optional implementation, in the first aspect of the present invention, the prediction network model is a classification algorithm model; before determining a plurality of candidate listing products, the method further includes:

[0029] The multi-level sample training set corresponding to the candidate shelf goods is input into the classification algorithm training model for training to obtain the classification algorithm model corresponding to the candidate shelf goods; the multi-level sample training set includes store addresses with multiple sales effect levels and corresponding store feature parameters.

[0030] As an optional implementation, in the first aspect of the present invention, the step of inputting the plurality of candidate listing products and the corresponding store characteristic parameters into the prediction network model corresponding to each of the candidate listing products to determine the target candidate listing product comprises:

[0031] For any of the candidate products on the shelves, the candidate products on the shelves and the corresponding store feature parameters are input into the classification algorithm model corresponding to the candidate products on the shelves, and the classification result of the classification algorithm model on the candidate products on the shelves is output;

[0032] Filter out candidate listing products that are classified as having excellent sales effects from all the candidate listing products;

[0033] According to the candidate listing products classified as having excellent sales effects, target candidate listing products are determined.

[0034] A second aspect of an embodiment of the present invention discloses a device for determining shelf goods, the device comprising:

[0035] A product determination module is used to determine multiple candidate products for listing;

[0036] A parameter calculation module, used to calculate the store characteristic parameters of the store corresponding to each of the candidate listing products;

[0037] The product prediction module is used to input the multiple candidate shelves of goods and the corresponding store characteristic parameters into the prediction network model corresponding to each of the candidate shelves of goods, so as to determine the target candidate shelves of goods; the prediction network model is obtained by training a training set including multiple stores with known sales effects and corresponding store characteristic parameters of the corresponding candidate shelves of goods.

[0038] As an optional implementation, in the second aspect of the present invention, the store characteristic parameters include at least one of store surrounding traffic characteristics, store surrounding facility characteristics and store surrounding competitive product characteristics.

[0039] As an optional implementation, in the second aspect of the present invention, the specific manner in which the parameter calculation module calculates the store characteristic parameters of each of the candidate listing products includes:

[0040] Determine the surrounding traffic information of the listed store;

[0041] Determine the surrounding crowd flow portrait corresponding to the candidate listing product according to the surrounding crowd flow information and a preset crowd portrait matching database;

[0042] Determine the surrounding crowd flow portrait as the surrounding crowd flow feature of the store of the candidate product;

[0043] and / or,

[0044] Determine the surrounding traffic information of the listed store;

[0045] Determine a number of target facilities and a number of competing stores corresponding to the listed store;

[0046] Determine the surrounding traffic information of the target facility and the competing store;

[0047] Calculate the crowd flow overlap between the surrounding crowd flow information of the listed store and the surrounding crowd flow information of any of the target facilities or the competing stores;

[0048] The overlap of all the crowds is determined as the crowd flow characteristics around the store of the candidate listing product.

[0049] As an optional implementation, in the second aspect of the present invention, the specific manner in which the parameter calculation module calculates the store characteristic parameters of each of the candidate listing products includes:

[0050] Determine a plurality of target facilities around the location of the listed store;

[0051] Determine a distance parameter between the location of the listed store and any of the target facilities;

[0052] The distance parameters corresponding to all the target facilities are determined as the store surrounding facility features of the candidate shelf products.

[0053] As an optional implementation, in the second aspect of the present invention, the specific manner in which the parameter calculation module calculates the store characteristic parameters of each of the candidate listing products includes:

[0054] Determine multiple competing stores around the location of the listed store;

[0055] Determine a distance parameter between the location of the listing store and any of the competing stores;

[0056] The distance parameters corresponding to all the competing product stores are determined as the competing product features around the stores of the candidate listed products.

[0057] As an optional implementation, in the second aspect of the present invention, the surrounding human flow information is device information of all users within the coverage area of ​​the corresponding human flow.

[0058] As an optional implementation, in the second aspect of the present invention, the prediction network model is a classification algorithm model; the device further includes:

[0059] The training module is used to input the multi-level sample training set corresponding to the candidate shelf goods into the classification algorithm training model for training to obtain the classification algorithm model corresponding to the candidate shelf goods; the multi-level sample training set includes store addresses with multiple sales effect levels and corresponding store feature parameters.

[0060] As an optional implementation, in the second aspect of the present invention, the product prediction module inputs the multiple candidate shelves and the corresponding store feature parameters into the prediction network model corresponding to each of the candidate shelves to determine the specific method of the target candidate shelves, including:

[0061] For any of the candidate products on the shelves, the candidate products on the shelves and the corresponding store feature parameters are input into the classification algorithm model corresponding to the candidate products on the shelves, and the classification result of the classification algorithm model on the candidate products on the shelves is output;

[0062] Filter out candidate listing products that are classified as having excellent sales effects from all the candidate listing products;

[0063] According to the candidate listing products classified as having excellent sales effects, target candidate listing products are determined.

[0064] The third aspect of the present invention discloses another device for determining shelf goods, the device comprising:

[0065] A memory storing executable program code;

[0066] a processor coupled to the memory;

[0067] The processor calls the executable program code stored in the memory to execute part or all of the steps in the method for determining shelf goods disclosed in the first aspect of the present invention.

[0068] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0069] In an embodiment of the present invention, a method and device for determining shelf goods are disclosed, the method comprising: determining a plurality of candidate shelf goods; calculating the store characteristic parameters of the shelf store corresponding to each of the candidate shelf goods; inputting the plurality of candidate shelf goods and the corresponding store characteristic parameters into the prediction network model corresponding to each of the candidate shelf goods, respectively, to determine the target candidate shelf goods; the prediction network model is obtained by training with a training set including stores with known sales effects of the plurality of corresponding candidate shelf goods and the corresponding store characteristic parameters. It can be seen that the embodiment of the present invention can use the trained prediction network model to calculate and predict the plurality of candidate shelf goods and the corresponding store characteristic parameters, to determine the target candidate shelf goods, so as to be able to combine artificial intelligence technology to assist in the selection of shelf goods. On the one hand, it can fully combine the store characteristic parameters of the store to consider the selection of goods for listing, and on the other hand, it achieves an efficient and accurate effect of determining the listing of goods. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0071] Figure 1 It is a flow chart of a method for determining shelf goods disclosed in an embodiment of the present invention.

[0072] Figure 2 It is a structural schematic diagram of a device for determining shelf goods disclosed in an embodiment of the present invention.

[0073] Figure 3 It is a structural schematic diagram of another device for determining shelf goods disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0074] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or equipment that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or equipment.

[0076] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0077] The present invention discloses a method and device for determining goods to be put on the shelves, which can use a trained prediction network model to calculate and predict multiple candidate goods to be put on the shelves and corresponding store characteristic parameters to determine the target candidate goods to be put on the shelves, so that artificial intelligence technology can be combined to assist in selecting goods to be put on the shelves. On the one hand, the store characteristic parameters of the store can be fully combined to consider the selection of goods for putting on the shelves, and on the other hand, an efficient and accurate effect of determining the goods to be put on the shelves is achieved. The following are detailed descriptions.

[0078] Embodiment 1

[0079] See also Figure 1 , Figure 1 1 is a flow chart of a method for determining shelf goods disclosed in an embodiment of the present invention. Figure 1 The described method for determining the shelf goods is applied to a determination chip, a determination terminal or a determination server (wherein the determination server may be a local server or a cloud server) of a store shelf goods selection system. Figure 1 As shown, the method for determining the shelf goods may include the following operations:

[0080] 101. Identify multiple candidate products for listing.

[0081] Optionally, the candidate items for listing can be obtained by capturing them from a preset list of candidate items for listing, or they can be manually input through a terminal. For example, an operator can determine several potential candidate items for listing after performing some simple data analysis, and input these items through a terminal to form multiple candidate items for listing.

[0082] 102. Calculate the store characteristic parameters of the store corresponding to each candidate listing product.

[0083] In the embodiment of the present invention, the listed store can be any type of store, such as grocery stores, convenience stores, discount stores, supermarkets, hypermarkets, warehouse membership stores, department stores, specialty stores, specialty stores, furniture and building materials stores, shopping centers including community shopping centers, urban shopping centers, suburban shopping centers or factory direct sales centers and other types of stores.

[0084] In an embodiment of the present invention, the store characteristic parameters include at least one of the following: store surrounding traffic characteristics, store surrounding facility characteristics, and store surrounding competitive product characteristics. Optionally, the store surrounding traffic characteristics may include one or more of the following information: a store surrounding traffic portrait, a store surrounding traffic overlap degree, and surrounding competitive product stores.

[0085] 103. Input a plurality of candidate listing products and corresponding store characteristic parameters into a prediction network model corresponding to each candidate listing product to determine a target candidate listing product.

[0086] In the embodiment of the present invention, the prediction network model is obtained by training a training set including a plurality of stores with known sales effects of corresponding candidate listing products and corresponding store characteristic parameters. In the embodiment of the present invention, the target candidate listing products can be used for listing for sale or for further product evaluation.

[0087] It can be seen that the above-mentioned embodiments of the invention can use the trained prediction network model to calculate and predict multiple candidate shelves of goods and corresponding store characteristic parameters to determine the target candidate shelves of goods, so as to combine artificial intelligence technology to assist in the selection of shelves of goods. On the one hand, it can fully combine the store characteristic parameters of the store to consider the selection of goods for listing, and on the other hand, it achieves efficient and accurate goods listing determination effect.

[0088] As an optional implementation, in step 102, calculating the store characteristic parameters of the store corresponding to each candidate listing product includes:

[0089] Determine the surrounding traffic information of the store;

[0090] According to the surrounding crowd flow information and the preset crowd portrait matching database, determine the surrounding crowd flow portrait corresponding to the candidate listing products;

[0091] The surrounding traffic profile is determined as the surrounding traffic characteristics of the store where the candidate products are put on the shelves.

[0092] In an embodiment of the present invention, the surrounding crowd flow information may be the device information or image information of all users within the corresponding crowd flow coverage area. Optionally, the device information may be a device ID, and the image information may be a facial image or a full-body image of the user. Optionally, the crowd portrait matching database may include a correspondence between the device information or image information of multiple users and the user's portrait features, wherein the user's portrait features may include but are not limited to the user's gender, occupation, hobbies, specialties, interests, age and other information. Optionally, based on the surrounding crowd flow information and a preset crowd portrait matching database, determining the surrounding crowd flow portrait corresponding to the candidate listing product may include:

[0093] For the device information or image information of any user in the surrounding crowd flow information, the corresponding portrait features are matched according to the crowd portrait matching database;

[0094] According to the portrait features corresponding to all users in the surrounding traffic information, the surrounding traffic portrait corresponding to the candidate listing products is determined.

[0095] Optionally, the portrait features corresponding to all users in the surrounding crowd flow information can be directly determined as the surrounding crowd flow portraits corresponding to the candidate listing products, or the portrait features corresponding to all users in the surrounding crowd flow information can be statistically analyzed in different categories or ranked in different categories to obtain statistical data corresponding to the portrait features corresponding to all users, and the statistical data corresponding to the portrait features corresponding to all users can be determined as the surrounding crowd flow portraits corresponding to the candidate listing products.

[0096] It can be seen that through this optional implementation method, the surrounding crowd flow portrait corresponding to the candidate shelf goods can be calculated, and the surrounding crowd flow portrait can be determined as the store characteristic parameters of the candidate shelf goods, so as to reasonably determine the crowd flow characteristics around the store, which is conducive to the subsequent use of the trained prediction network model to calculate and predict multiple candidate shelf goods and the corresponding store characteristic parameters to determine the target candidate shelf goods.

[0097] As an optional implementation, in step 102, calculating the store characteristic parameters of the store corresponding to each candidate listing product includes:

[0098] Determine the surrounding traffic information of the store;

[0099] Determine a number of target facilities and a number of competing stores corresponding to the store listing;

[0100] Determine the surrounding traffic information of target facilities and competing stores;

[0101] Calculate the overlap between the surrounding traffic information of the listed store and the surrounding traffic information of any target facility or competing store;

[0102] The overlap of all traffic flows is determined as the traffic flow characteristics around the store where the candidate products are put on the shelves.

[0103] In an embodiment of the present invention, the surrounding crowd flow information may be the device information of all users within the corresponding crowd flow coverage. Specifically, the surrounding crowd flow information may be determined by determining a calculation object, such as a crowd flow coverage of a listed store, a target facility, or a competing store, and then further calculating the device information of all users within the crowd flow coverage area in the area corresponding to the crowd flow coverage area. For example, the device information and corresponding location information of all users in the area may be obtained, and the device information of all users within the crowd flow coverage area may be filtered out according to the location information. Optionally, the device information may be a device ID.

[0104] Optionally, the crowd flow overlap can be determined by calculating the overlap of all user device information in the surrounding crowd flow information of the listed store and all user device information in the surrounding crowd flow information of any target facility or competing store. For example, the crowd flow overlap can be determined as the ratio of the number of identical user device information among all user device information in the surrounding crowd flow information of the listed store and all user device information in the surrounding crowd flow information of any target facility or competing store, to the total number of all user device information in the surrounding crowd flow information of the listed store.

[0105] In an embodiment of the present invention, the surrounding crowd flow information may be the image information of all users within the corresponding crowd flow coverage area. Specifically, the surrounding crowd flow information may be determined by determining the calculation object, such as the crowd flow coverage area of ​​the listed store, target facility, or competitor store, and then further calculating the image information of all users within the crowd flow coverage area in the area corresponding to the crowd flow coverage area. For example, the monitoring information within the area may be obtained, and then the facial image information and corresponding location information of all users within the area may be identified through a facial image recognition algorithm, and the facial image information of all users within the crowd flow coverage area may be screened out based on the location information.

[0106] Optionally, the crowd flow overlap can be determined by calculating the overlap of all user image information in the surrounding crowd flow information of the listed store and all user image information in the surrounding crowd flow information of any target facility or competing store. For example, the crowd flow overlap can be determined as the similarity between all user image information in the surrounding crowd flow information of the listed store and all user image information in the surrounding crowd flow information of any target facility or competing store.

[0107] Optionally, the pedestrian coverage range can be determined according to the customer flow influence of the calculation object, which can be an empirical value. For example, the pedestrian coverage range of a listed store or a competing store is generally set to a circle with a radius of 100 meters with the store location as the center, and the pedestrian coverage range of a target facility such as a shopping mall is generally set to a circle with a radius of 1000 meters with the store location as the center, and the pedestrian coverage range of a target facility such as an office building is generally set to a circle with a radius of 300 meters with the store location as the center.

[0108] It can be seen that through this optional implementation method, the crowd flow overlap between the surrounding crowd flow information of the listed store and the surrounding crowd flow information of any target facility or competing store can be calculated, and all crowd flow overlaps can be determined as the store characteristic parameters of the candidate listed goods, thereby reasonably determining the crowd flow characteristics around the store, which is conducive to the subsequent use of the trained prediction network model to calculate and predict multiple candidate listed goods and the corresponding store characteristic parameters, so as to determine the target candidate listed goods.

[0109] As an optional implementation, in step 102, calculating the store characteristic parameters of the store corresponding to each candidate listing product includes:

[0110] Identify multiple target facilities around the location of the listed store;

[0111] Determine the distance parameters between the location of the listed store and any target facility;

[0112] The distance parameters corresponding to all target facilities are determined as the store surrounding facility features of the candidate listing products.

[0113] In the embodiment of the present invention, optionally, the target facility corresponding to the listed store can be a commercial facility within the first distance range of the location of the listed store, such as a shopping mall or office building, such as a clothing, department store, building materials, decorative materials market or a comprehensive shopping mall, or other large-scale catering, entertainment, leisure facilities, commercial plazas and commercial streets, and other iconic commercial facilities closely related to people's lives.

[0114] Optionally, the distance parameter between the location of the listed store and any target facility may include one or both of a straight-line distance and a walkable distance. Optionally, the straight-line distance between the location of the listed store and any target facility may be determined by calculating the distance between the location of the listed store and the location of any target facility. Optionally, the walkable distance between the location of the listed store and any target facility may be determined by the following method:

[0115] According to a preset regional map model, a walking path between the location of the listed store and the location of any target facility in the regional map model is determined;

[0116] The length of the walking path is determined to determine the walkable distance between the location of the listed store and any target facility.

[0117] It can be seen that through this optional implementation method, the distance parameters between the location of the listed store and any target facility can be determined, and the distance parameters corresponding to all target facilities can be determined as the store surrounding facility characteristics of the candidate listed goods, thereby reasonably determining the store surrounding facility characteristics, which is conducive to the subsequent use of the trained prediction network model to calculate and predict multiple candidate listed goods and corresponding store characteristic parameters to determine the target candidate listed goods.

[0118] As an optional implementation, in step 102, calculating the store characteristic parameters of the store corresponding to each candidate listing product includes:

[0119] Determine multiple competing stores around the location of the listed store;

[0120] Determine a distance parameter between the location of the listing store and any of the competing stores;

[0121] The distance parameters corresponding to all the competing product stores are determined as the competing product features around the stores of the candidate listed products.

[0122] In the embodiment of the present invention, optionally, the competing store corresponding to the listing store may be a competing store around the location of the listing store, wherein the competing store may be a surrounding store whose service field intersects with the service field of the store corresponding to the candidate listing product, or a surrounding store whose product parameters of the listing product intersect with the product parameters of the preset listing product corresponding to the candidate listing product. Optionally, the method for determining the competing store may include:

[0123] Get all stores within the second distance range of the listed store's location;

[0124] Determine the store parameters corresponding to any store; the store parameters include a service area set and / or product parameters of the listed products;

[0125] Determine the store parameters corresponding to the listed store;

[0126] Calculate the similarity between the store parameters corresponding to the listed store and the store parameters corresponding to any store;

[0127] Stores whose similarity between store parameters corresponding to the listed stores is higher than a preset similarity threshold are determined as competing stores.

[0128] Optionally, the distance parameter between the location of the listed store and any competing store may include one or both of a straight-line distance and a walkable distance. Optionally, the straight-line distance between the location of the listed store and any competing store may be determined by calculating the distance between the location of the listed store and the location of any competing store. Optionally, the walkable distance between the location of the listed store and any competing store may be determined by the following method:

[0129] According to the preset regional map model, determine the walking path between the location of the listed store and the location of any competing store in the regional map model;

[0130] Determine the length of this walking path to determine the walking distance between the listing store's location and any competing stores.

[0131] It can be seen that through this optional implementation method, the distance parameter between the location of the listed store and any competing store can be determined, and the distance parameters corresponding to all competing stores can be determined as the store surrounding facility characteristics of the candidate listed goods, thereby reasonably determining the store surrounding facility characteristics, which is conducive to the subsequent use of the trained prediction network model to calculate and predict multiple candidate listed goods and the corresponding store characteristic parameters, so as to determine the target candidate listed goods.

[0132] As an optional implementation, the prediction network model is a classification algorithm model, for example, it can be a neural network classification algorithm model, such as a Bayesian classification model or a decision tree model, or other algorithm models that can be used for classification.

[0133] Specifically, before determining a plurality of candidate products for listing in step 101, the method further includes:

[0134] The multi-level sample training set corresponding to the candidate listing items is input into the classification algorithm training model for training to obtain the classification algorithm model corresponding to the candidate listing items.

[0135] In an embodiment of the present invention, a multi-level sample training set includes store addresses of multiple sales effect levels and corresponding store feature parameters. Optionally, the multi-level sample training set may be a positive and negative sample training set, which includes multiple high-quality store addresses and corresponding store feature parameters, and multiple low-quality store addresses and corresponding store feature parameters. Optionally, a high-quality store address is the address information of a store with a good sales record for selling the candidate product on the shelf, and correspondingly, a low-quality store address is the address information of a store with a poor sales record for selling the candidate product on the shelf. Optionally, the method for determining a high-quality store address or a low-quality store address may include:

[0136] Get multiple historical store addresses;

[0137] Determine the sales volume of the candidate listing product in the historical time period at the store corresponding to each historical store address;

[0138] All historical items on the shelves with sales higher than a preset sales threshold are determined as high-quality store addresses, and all historical items on the shelves with sales lower than a preset sales threshold are determined as low-quality store addresses.

[0139] In an embodiment of the present invention, multiple high-quality store addresses and corresponding store feature parameters in a multi-level sample training set can be used as a training set marked as excellent, and multiple low-quality store addresses and corresponding store feature parameters can be used as a training set marked as poor. The multi-level sample training set is input into a classification algorithm training model for training until the classification algorithm training model converges, and a classification algorithm model can be obtained. The classification algorithm model obtained by training can be used to predict the input candidate shelf goods and the corresponding store feature parameters to obtain a prediction score, and when the prediction score is higher than the prediction threshold obtained by training, the candidate shelf goods are determined to have an excellent sales effect.

[0140] Optionally, the multi-level sample training set may also be a three-level sample training set, which includes multiple high-quality store addresses and corresponding store feature parameters, multiple general store addresses and corresponding store feature parameters, and multiple low-quality store addresses and corresponding store feature parameters. Optionally, the high-quality store address is the address information of a store with a good sales record for selling the candidate goods on the shelves, and correspondingly, the general store address is the address information of a store with a general sales record for selling the candidate goods on the shelves, and the low-quality store address is the address information of a store with a poor sales record for selling the candidate goods on the shelves. Optionally, the method for determining the high-quality store address, the general store address, or the low-quality store address may include:

[0141] Get multiple historical store addresses;

[0142] Determine the sales volume of the candidate listing product in the historical time period at the store corresponding to each historical store address;

[0143] All historically-stored goods whose sales are within the preset high-quality sales range are determined as high-quality store addresses, all historically-stored goods whose sales are within the preset general sales range are determined as general store addresses, and all historically-stored goods whose sales are within the preset low-quality sales range are determined as low-quality store addresses.

[0144] In an embodiment of the present invention, multiple high-quality store addresses and corresponding store feature parameters in a multi-level sample training set can be used as a training set marked as excellent, and multiple general store addresses and corresponding store feature parameters can be used as a training set marked as general, and multiple low-quality store addresses and corresponding store feature parameters can be used as a training set marked as poor. The multi-level sample training set is input into the classification algorithm training model for training until the classification algorithm training model converges, and then a classification algorithm model can be obtained. The classification algorithm model obtained by training can be used to predict the input candidate shelf goods and the corresponding store feature parameters to obtain a prediction score, and when the prediction score is higher than the prediction threshold obtained by training, the candidate shelf goods are determined to have an excellent sales effect.

[0145] Optionally, based on the above examples, the multi-level sample training set can also be a four-level sample training set or a five-level sample training set. The definition and determination methods can refer to the above examples. In fact, any labeling method of the sample training set that can be used to achieve classification purposes should be considered to be included in the scope of protection of the present invention.

[0146] It can be seen that through this optional implementation method, the multi-level sample training set corresponding to the candidate shelf goods can be input into the classification algorithm training model for training to obtain the classification algorithm model corresponding to the candidate shelf goods, so as to train the classification algorithm model, which is conducive to the subsequent use of the trained classification algorithm model to calculate and predict multiple candidate shelf goods and corresponding store feature parameters to determine the target candidate shelf goods.

[0147] As an optional implementation, in step 103, a plurality of candidate listing products and corresponding store characteristic parameters are respectively input into a prediction network model corresponding to each candidate listing product to determine a target candidate listing product, including:

[0148] For any candidate product, the candidate product and the corresponding store feature parameters are input into the classification algorithm model corresponding to the candidate product, and the classification result of the classification algorithm model for the candidate product is output;

[0149] Filter out candidate listing products that are classified as having excellent sales effects from all candidate listing products;

[0150] Based on the candidate listing products classified as having excellent sales effects, target candidate listing products are determined.

[0151] Optionally, target candidate shelf goods are determined based on candidate shelf goods classified as having excellent sales effects. The predicted scores corresponding to multiple candidate shelf goods classified as having excellent sales effects can be used to determine candidate shelf goods with predicted scores higher than a preset score threshold or the top preset number of places in the predicted score ranking as target candidate shelf goods.

[0152] It can be seen that through this optional implementation method, the target candidate goods for listing can be determined based on multiple candidate goods for listing that are classified as having excellent sales effects, which is conducive to further screening out the target candidate goods for listing and achieving efficient and accurate goods listing selection effects.

[0153] Embodiment 2

[0154] See also Figure 2 , Figure 2 Schematic diagram of a device for determining shelf goods disclosed in an embodiment of the present invention. Figure 2 The described device for determining the goods to be put on the shelves is applied to a determination chip, a determination terminal or a determination server (wherein the determination server may be a local server or a cloud server) of a store goods selection system. Figure 2 As shown, the device for determining the goods on the shelves may include:

[0155] The product determination module 201 is used to determine a plurality of candidate products to be put on the shelves.

[0156] Optionally, the candidate items for listing can be obtained by capturing them from a preset list of candidate items for listing, or they can be manually input through a terminal. For example, an operator can determine several potential candidate items for listing after performing some simple data analysis, and input these items through a terminal to form multiple candidate items for listing.

[0157] The parameter calculation module 202 is used to calculate the store characteristic parameters of the store corresponding to each candidate listing product.

[0158] In the embodiment of the present invention, the listed store can be any type of store, such as grocery stores, convenience stores, discount stores, supermarkets, hypermarkets, warehouse membership stores, department stores, specialty stores, specialty stores, furniture and building materials stores, shopping centers including community shopping centers, urban shopping centers, suburban shopping centers or factory direct sales centers and other types of stores.

[0159] In an embodiment of the present invention, the store characteristic parameters include at least one of the following: store surrounding traffic characteristics, store surrounding facility characteristics, and store surrounding competitive product characteristics. Optionally, the store surrounding traffic characteristics may include one or more of the following information: a store surrounding traffic portrait, a store surrounding traffic overlap degree, and surrounding competitive product stores.

[0160] The product prediction module 203 is used to input a plurality of candidate products to be put on the shelves and corresponding store characteristic parameters into the prediction network model corresponding to each candidate product to determine the target candidate product to be put on the shelves.

[0161] In the embodiment of the present invention, the prediction network model is obtained by training a training set including a plurality of stores with known sales effects of corresponding candidate listing products and corresponding store characteristic parameters. In the embodiment of the present invention, the target candidate listing products can be used for listing for sale or for further product evaluation.

[0162] It can be seen that the above-mentioned embodiments of the invention can use the trained prediction network model to calculate and predict multiple candidate shelves of goods and corresponding store characteristic parameters to determine the target candidate shelves of goods, so as to combine artificial intelligence technology to assist in the selection of shelves of goods. On the one hand, it can fully combine the store characteristic parameters of the store to consider the selection of goods for listing, and on the other hand, it achieves efficient and accurate goods listing determination effect.

[0163] As an optional implementation, the parameter calculation module 202 calculates the store characteristic parameters of each candidate listing store corresponding to each listing product in a specific manner, including:

[0164] Determine the surrounding traffic information of the store;

[0165] According to the surrounding crowd flow information and the preset crowd portrait matching database, determine the surrounding crowd flow portrait corresponding to the candidate listing products;

[0166] The surrounding traffic profile is determined as the surrounding traffic characteristics of the store where the candidate products are put on the shelves.

[0167] In an embodiment of the present invention, the surrounding crowd flow information may be the device information or image information of all users within the corresponding crowd flow coverage area. Optionally, the device information may be a device ID, and the image information may be a facial image or a full-body image of the user. Optionally, the crowd portrait matching database may include a correspondence between the device information or image information of multiple users and the user's portrait features, wherein the user's portrait features may include but are not limited to the user's gender, occupation, hobbies, specialties, interests, age and other information. Optionally, based on the surrounding crowd flow information and a preset crowd portrait matching database, determining the surrounding crowd flow portrait corresponding to the candidate listing product may include:

[0168] For the device information or image information of any user in the surrounding crowd flow information, the corresponding portrait features are matched according to the crowd portrait matching database;

[0169] According to the portrait features corresponding to all users in the surrounding traffic information, the surrounding traffic portrait corresponding to the candidate listing products is determined.

[0170] Optionally, the portrait features corresponding to all users in the surrounding crowd flow information can be directly determined as the surrounding crowd flow portraits corresponding to the candidate listing products, or the portrait features corresponding to all users in the surrounding crowd flow information can be statistically analyzed in different categories or ranked in different categories to obtain statistical data corresponding to the portrait features corresponding to all users, and the statistical data corresponding to the portrait features corresponding to all users can be determined as the surrounding crowd flow portraits corresponding to the candidate listing products.

[0171] It can be seen that through this optional implementation method, the surrounding crowd flow portrait corresponding to the candidate shelf goods can be calculated, and the surrounding crowd flow portrait can be determined as the store characteristic parameters of the candidate shelf goods, so as to reasonably determine the crowd flow characteristics around the store, which is conducive to the subsequent use of the trained prediction network model to calculate and predict multiple candidate shelf goods and the corresponding store characteristic parameters to determine the target candidate shelf goods.

[0172] As an optional implementation, the parameter calculation module 202 calculates the store characteristic parameters of each candidate listing store corresponding to each listing product in a specific manner, including:

[0173] Determine the surrounding traffic information of the store;

[0174] Determine a number of target facilities and a number of competing stores corresponding to the store listing;

[0175] Determine the surrounding traffic information of target facilities and competing stores;

[0176] Calculate the overlap between the surrounding traffic information of the listed store and the surrounding traffic information of any target facility or competing store;

[0177] The overlap of all traffic flows is determined as the traffic flow characteristics around the store where the candidate products are put on the shelves.

[0178] In an embodiment of the present invention, the surrounding crowd flow information may be the device information of all users within the corresponding crowd flow coverage. Specifically, the surrounding crowd flow information may be determined by determining a calculation object, such as a crowd flow coverage of a listed store, a target facility, or a competing store, and then further calculating the device information of all users within the crowd flow coverage area in the area corresponding to the crowd flow coverage area. For example, the device information and corresponding location information of all users in the area may be obtained, and the device information of all users within the crowd flow coverage area may be filtered out according to the location information. Optionally, the device information may be a device ID.

[0179] Optionally, the crowd flow overlap can be determined by calculating the overlap of all user device information in the surrounding crowd flow information of the listed store and all user device information in the surrounding crowd flow information of any target facility or competing store. For example, the crowd flow overlap can be determined as the ratio of the number of identical user device information among all user device information in the surrounding crowd flow information of the listed store and all user device information in the surrounding crowd flow information of any target facility or competing store, to the total number of all user device information in the surrounding crowd flow information of the listed store.

[0180] In an embodiment of the present invention, the surrounding crowd flow information may be the image information of all users within the corresponding crowd flow coverage area. Specifically, the surrounding crowd flow information may be determined by determining the calculation object, such as the crowd flow coverage area of ​​the listed store, target facility, or competitor store, and then further calculating the image information of all users within the crowd flow coverage area in the area corresponding to the crowd flow coverage area. For example, the monitoring information within the area may be obtained, and then the facial image information and corresponding location information of all users within the area may be identified through a facial image recognition algorithm, and the facial image information of all users within the crowd flow coverage area may be screened out based on the location information.

[0181] Optionally, the crowd flow overlap can be determined by calculating the overlap of all user image information in the surrounding crowd flow information of the listed store and all user image information in the surrounding crowd flow information of any target facility or competing store. For example, the crowd flow overlap can be determined as the similarity between all user image information in the surrounding crowd flow information of the listed store and all user image information in the surrounding crowd flow information of any target facility or competing store.

[0182] Optionally, the pedestrian coverage range can be determined according to the customer flow influence of the calculation object, which can be an empirical value. For example, the pedestrian coverage range of a listed store or a competing store is generally set to a circle with a radius of 100 meters with the store location as the center, and the pedestrian coverage range of a target facility such as a shopping mall is generally set to a circle with a radius of 1000 meters with the store location as the center, and the pedestrian coverage range of a target facility such as an office building is generally set to a circle with a radius of 300 meters with the store location as the center.

[0183] It can be seen that through this optional implementation method, the crowd flow overlap between the surrounding crowd flow information of the listed store and the surrounding crowd flow information of any target facility or competing store can be calculated, and all crowd flow overlaps can be determined as the store characteristic parameters of the candidate listed goods, thereby reasonably determining the crowd flow characteristics around the store, which is conducive to the subsequent use of the trained prediction network model to calculate and predict multiple candidate listed goods and the corresponding store characteristic parameters, so as to determine the target candidate listed goods.

[0184] As an optional implementation, the parameter calculation module 202 calculates the store characteristic parameters of each candidate listing store corresponding to each listing product in a specific manner, including:

[0185] Identify multiple target facilities around the location of the listed store;

[0186] Determine the distance parameters between the location of the listed store and any target facility;

[0187] The distance parameters corresponding to all target facilities are determined as the store surrounding facility features of the candidate listing products.

[0188] In the embodiment of the present invention, optionally, the target facility corresponding to the listed store can be a commercial facility within the first distance range of the location of the listed store, such as a shopping mall or office building, such as a clothing, department store, building materials, decorative materials market or a comprehensive shopping mall, or other large-scale catering, entertainment, leisure facilities, commercial plazas and commercial streets, and other iconic commercial facilities closely related to people's lives.

[0189] Optionally, the distance parameter between the location of the listed store and any target facility may include one or both of a straight-line distance and a walkable distance. Optionally, the straight-line distance between the location of the listed store and any target facility may be determined by calculating the distance between the location of the listed store and the location of any target facility. Optionally, the walkable distance between the location of the listed store and any target facility may be determined by the following method:

[0190] According to a preset regional map model, a walking path between the location of the listed store and the location of any target facility in the regional map model is determined;

[0191] The length of the walking path is determined to determine the walkable distance between the location of the listed store and any target facility.

[0192] It can be seen that through this optional implementation method, the distance parameters between the location of the listed store and any target facility can be determined, and the distance parameters corresponding to all target facilities can be determined as the store surrounding facility characteristics of the candidate listed goods, thereby reasonably determining the store surrounding facility characteristics, which is conducive to the subsequent use of the trained prediction network model to calculate and predict multiple candidate listed goods and corresponding store characteristic parameters to determine the target candidate listed goods.

[0193] As an optional implementation, the parameter calculation module 202 calculates the store characteristic parameters of each candidate listing store corresponding to each listing product in a specific manner, including:

[0194] Identify multiple competing stores around the location of the listed store;

[0195] Determine the distance parameter between the location of the listed store and any competing store;

[0196] The distance parameters corresponding to all competing stores are determined as the competing product features around the stores of the candidate listing products.

[0197] In the embodiment of the present invention, optionally, the competing store corresponding to the listing store may be a competing store around the location of the listing store, wherein the competing store may be a surrounding store whose service field intersects with the service field of the store corresponding to the candidate listing product, or a surrounding store whose product parameters of the listing product intersect with the product parameters of the preset listing product corresponding to the candidate listing product. Optionally, the method for determining the competing store may include:

[0198] Get all stores within the second distance range of the listed store's location;

[0199] Determine the store parameters corresponding to any store; the store parameters include a service area set and / or product parameters of the listed products;

[0200] Determine the store parameters corresponding to the listed store;

[0201] Calculate the similarity between the store parameters corresponding to the listed store and the store parameters corresponding to any store;

[0202] Stores whose similarity between store parameters corresponding to the listed stores is higher than a preset similarity threshold are determined as competing stores.

[0203] Optionally, the distance parameter between the location of the listed store and any competing store may include one or both of a straight-line distance and a walkable distance. Optionally, the straight-line distance between the location of the listed store and any competing store may be determined by calculating the distance between the location of the listed store and the location of any competing store. Optionally, the walkable distance between the location of the listed store and any competing store may be determined by the following method:

[0204] According to the preset regional map model, determine the walking path between the location of the listed store and the location of any competing store in the regional map model;

[0205] Determine the length of this walking path to determine the walking distance between the listing store's location and any competing stores.

[0206] It can be seen that through this optional implementation method, the distance parameter between the location of the listed store and any competing store can be determined, and the distance parameters corresponding to all competing stores can be determined as the store surrounding facility characteristics of the candidate listed goods, thereby reasonably determining the store surrounding facility characteristics, which is conducive to the subsequent use of the trained prediction network model to calculate and predict multiple candidate listed goods and the corresponding store characteristic parameters, so as to determine the target candidate listed goods.

[0207] As an optional implementation, the prediction network model is a classification algorithm model, for example, it can be a neural network classification algorithm model, such as a Bayesian classification model or a decision tree model, or other algorithm models that can be used for classification.

[0208] Specifically, the device also includes:

[0209] The training module is used to input the multi-level sample training set corresponding to the candidate shelf goods into the classification algorithm training model for training to obtain the classification algorithm model corresponding to the candidate shelf goods.

[0210] In an embodiment of the present invention, the multi-level sample training set includes store addresses of multiple sales effect levels and corresponding store characteristic parameters. Optionally, the high-quality store address is the address information of a store with a good sales record for selling the candidate product on the shelf, and correspondingly, the low-quality store address is the address information of a store with a poor sales record for selling the candidate product on the shelf. Optionally, the method for determining the high-quality store address or the low-quality store address may include:

[0211] Get multiple historical store addresses;

[0212] Determine the sales volume of the candidate listing product in the store corresponding to each historical store address during the historical time period;

[0213] All historical items on the shelves with sales higher than a preset sales threshold are determined as high-quality store addresses, and all historical items on the shelves with sales lower than a preset sales threshold are determined as low-quality store addresses.

[0214] In an embodiment of the present invention, multiple high-quality store addresses and corresponding store feature parameters in a multi-level sample training set can be used as a training set marked as excellent, and multiple low-quality store addresses and corresponding store feature parameters can be used as a training set marked as poor. The multi-level sample training set is input into a classification algorithm training model for training until the classification algorithm training model converges, and a classification algorithm model can be obtained. The classification algorithm model obtained by training can be used to predict the input candidate shelf goods and the corresponding store feature parameters to obtain a prediction score, and when the prediction score is higher than the prediction threshold obtained by training, the candidate shelf goods are determined to have an excellent sales effect.

[0215] Optionally, the multi-level sample training set may also be a three-level sample training set, which includes multiple high-quality store addresses and corresponding store feature parameters, multiple general store addresses and corresponding store feature parameters, and multiple low-quality store addresses and corresponding store feature parameters. Optionally, the high-quality store address is the address information of a store with a good sales record for selling the candidate goods on the shelves, and correspondingly, the general store address is the address information of a store with a general sales record for selling the candidate goods on the shelves, and the low-quality store address is the address information of a store with a poor sales record for selling the candidate goods on the shelves. Optionally, the method for determining the high-quality store address, the general store address, or the low-quality store address may include:

[0216] Get multiple historical store addresses;

[0217] Determine the sales volume of the candidate listing product in the historical time period at the store corresponding to each historical store address;

[0218] All historically-stored goods whose sales are within the preset high-quality sales range are determined as high-quality store addresses, all historically-stored goods whose sales are within the preset general sales range are determined as general store addresses, and all historically-stored goods whose sales are within the preset low-quality sales range are determined as low-quality store addresses.

[0219] In an embodiment of the present invention, multiple high-quality store addresses and corresponding store feature parameters in a multi-level sample training set can be used as a training set marked as excellent, and multiple general store addresses and corresponding store feature parameters can be used as a training set marked as general, and multiple low-quality store addresses and corresponding store feature parameters can be used as a training set marked as poor. The multi-level sample training set is input into the classification algorithm training model for training until the classification algorithm training model converges, and then a classification algorithm model can be obtained. The classification algorithm model obtained by training can be used to predict the input candidate shelf goods and the corresponding store feature parameters to obtain a prediction score, and when the prediction score is higher than the prediction threshold obtained by training, the candidate shelf goods are determined to have an excellent sales effect.

[0220] Optionally, based on the above examples, the multi-level sample training set can also be a four-level sample training set or a five-level sample training set. The definition and determination methods can refer to the above examples. In fact, any labeling method of the sample training set that can be used to achieve classification purposes should be considered to be included in the scope of protection of the present invention.

[0221] It can be seen that through this optional implementation method, the multi-level sample training set corresponding to the candidate shelf goods can be input into the classification algorithm training model for training to obtain the classification algorithm model corresponding to the candidate shelf goods, so as to train the classification algorithm model, which is conducive to the subsequent use of the trained classification algorithm model to calculate and predict multiple candidate shelf goods and corresponding store feature parameters to determine the target candidate shelf goods.

[0222] As an optional implementation, in the second aspect of the present invention, the product prediction module 203 inputs a plurality of candidate products and corresponding store feature parameters into the prediction network model corresponding to each candidate product to determine the specific method of the target candidate product, including:

[0223] For any candidate product, the candidate product and the corresponding store feature parameters are input into the classification algorithm model corresponding to the candidate product, and the classification result of the classification algorithm model for the candidate product is output;

[0224] Filter out candidate listing products that are classified as having excellent sales effects from all candidate listing products;

[0225] Based on the candidate listing products classified as having excellent sales effects, target candidate listing products are determined.

[0226] Optionally, target candidate shelf goods may be determined based on multiple candidate shelf goods classified as having excellent sales effects. The predicted scores corresponding to the multiple candidate shelf goods classified as having excellent sales effects may be used to determine the target candidate shelf goods as candidate shelf goods whose predicted scores are higher than a preset score threshold or the top preset number of places in the predicted score ranking.

[0227] It can be seen that through this optional implementation method, the target candidate goods for listing can be determined based on multiple candidate goods for listing that are classified as having excellent sales effects, which is conducive to further screening out the target candidate goods for listing and achieving efficient and accurate goods listing selection effects.

[0228] Embodiment 3

[0229] See also Figure 3 , Figure 3 It is another device for determining shelf goods disclosed in an embodiment of the present invention. Figure 3 The described device for determining the goods to be put on the shelves is applied to a determination chip, a determination terminal or a determination server (wherein the determination server may be a local server or a cloud server) of a store goods selection system. Figure 3 As shown, the device for determining the goods on the shelves may include:

[0230] A memory 301 storing executable program codes;

[0231] a processor 302 coupled to the memory 301;

[0232] The processor 302 calls the executable program code stored in the memory 301 to execute the steps of the method for determining the shelf goods described in the first or second embodiment.

[0233] Embodiment 4

[0234] An embodiment of the present invention discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps of the method for determining shelf goods described in the first or second embodiment.

[0235] Embodiment 5

[0236] An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute the steps of the method for determining shelf goods described in embodiment one or embodiment two.

[0237] The device embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, i.e., they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0238] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution can be essentially or partly contributed to the prior art in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, a magnetic disk storage, a magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0239] Finally, it should be noted that the method and device for determining shelf goods disclosed in the embodiments of the present invention disclose only the preferred embodiments of the present invention, which are only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for determining shelf goods, characterized in that: The method comprises: Identify multiple candidate products for listing; Calculate the store characteristic parameters of the store corresponding to each candidate listing product; Input the plurality of candidate listing products and the corresponding store characteristic parameters into the prediction network model corresponding to each of the candidate listing products to determine the target candidate listing product; the prediction network model is obtained by training a training set including stores with known sales effects of the plurality of corresponding candidate listing products and the corresponding store characteristic parameters; The store characteristic parameters include at least one of the following: store surrounding traffic characteristics, store surrounding facility characteristics, and store surrounding competitive product characteristics; The calculating of the store characteristic parameters of each candidate listing product includes: Determine a plurality of target facilities around the location of the listed store; Determine a distance parameter between the location of the listed store and any of the target facilities; Determine the distance parameters corresponding to all the target facilities as the store surrounding facility features of the candidate listing products; The distance parameter between the location of the listed store and any target facility includes one or both of a straight-line distance and a walkable distance, and the straight-line distance between the location of the listed store and any target facility is determined by calculating the distance between the location of the listed store and the location of any target facility; The walking distance between the location of the listed store and any target facility is determined as follows: According to a preset regional map model, determining a walking path between the location of the listed store and the location of any target facility in the regional map model; Determine the length of the walking path to determine the walkable distance between the location of the listed store and any target facility; The calculating of the store characteristic parameters of each candidate product on the shelves also includes: Determine the surrounding traffic information of the listed store; Determine the surrounding crowd flow portrait corresponding to the candidate listing product according to the surrounding crowd flow information and a preset crowd portrait matching database; Determine the surrounding crowd flow portrait as the surrounding crowd flow feature of the store of the candidate product; and / or, Determine the surrounding traffic information of the listed store; Determine a number of target facilities and a number of competing stores corresponding to the listed store; Determine the surrounding traffic information of the target facility and the competing store; Calculate the crowd flow overlap between the surrounding crowd flow information of the listed store and the surrounding crowd flow information of any of the target facilities or the competing stores; Determine the overlap of all the crowd flows as the crowd flow characteristics around the store of the candidate product; The surrounding crowd flow information is the device information or image information of all users within the corresponding crowd flow coverage area, the device information is the device ID, the image information is the user's face image or full-body image, and the crowd portrait matching database includes the corresponding relationship between the device information or image information of multiple users and the user's portrait features, and the user's portrait features include: the user's gender, occupation, hobby, specialty, interest, and age; The method for determining the surrounding pedestrian flow information is: calculating the pedestrian flow coverage of the listed store, the target facility or the competing store, and calculating the device information or image information of all users within the pedestrian flow coverage area in the area corresponding to the pedestrian flow coverage area, wherein the pedestrian flow coverage area is determined according to the passenger flow influence of the listed store, the target facility or the competing store; The crowd flow overlap is: the ratio of the number of identical user device information in the surrounding crowd flow information of the listed store and all user device information in the surrounding crowd flow information of any target facility or competing store to the total number of all user device information in the surrounding crowd flow information of the listed store; or The similarity between all user image information in the surrounding traffic information of the listed store and all user image information in the surrounding traffic information of any target facility or competing store.

2. The method for determining shelf goods according to claim 1, characterized in that: The calculating of the store characteristic parameters of each candidate listing product includes: Determine multiple competing stores around the location of the listed store; Determine a distance parameter between the location of the listing store and any of the competing stores; The distance parameters corresponding to all the competing product stores are determined as the competing product features around the stores of the candidate listed products.

3. The method for determining shelf goods according to claim 1, characterized in that: The prediction network model is a classification algorithm model; before determining a plurality of candidate products for listing, the method further includes: The multi-level sample training set corresponding to the candidate shelf goods is input into the classification algorithm training model for training to obtain the classification algorithm model corresponding to the candidate shelf goods; the multi-level sample training set includes store addresses with multiple sales effect levels and corresponding store feature parameters.

4. The method for determining shelf goods according to claim 1, characterized in that: The step of inputting the plurality of candidate listing products and the corresponding store characteristic parameters into the prediction network model corresponding to each of the candidate listing products to determine the target candidate listing products includes: For any of the candidate products on the shelves, the candidate products on the shelves and the corresponding store feature parameters are input into the classification algorithm model corresponding to the candidate products on the shelves, and the classification result of the classification algorithm model on the candidate products on the shelves is output; Filter out candidate listing products that are classified as having excellent sales effects from all the candidate listing products; According to the candidate listing products classified as having excellent sales effects, target candidate listing products are determined.

5. A device for determining shelf goods, characterized in that: The device comprises: A product determination module is used to determine multiple candidate products for listing; A parameter calculation module, used to calculate the store characteristic parameters of the store corresponding to each of the candidate listing products; A product prediction module, used to input the multiple candidate products and corresponding store characteristic parameters into the prediction network model corresponding to each of the candidate products, so as to determine the target candidate products; the prediction network model is obtained by training a training set including stores with known sales effects of the multiple corresponding candidate products and corresponding store characteristic parameters; The store characteristic parameters include at least one of the following: store surrounding traffic characteristics, store surrounding facility characteristics, and store surrounding competitive product characteristics; The specific method in which the parameter calculation module calculates the store characteristic parameters of each candidate listing product includes: Determine a plurality of target facilities around the location of the listed store; Determine a distance parameter between the location of the listed store and any of the target facilities; Determine the distance parameters corresponding to all the target facilities as the store surrounding facility features of the candidate listing products; The distance parameter between the location of the listed store and any target facility includes one or both of a straight-line distance and a walkable distance, and the straight-line distance between the location of the listed store and any target facility is determined by calculating the distance between the location of the listed store and the location of any target facility; The walking distance between the location of the listed store and any target facility is determined as follows: According to a preset regional map model, determining a walking path between the location of the listed store and the location of any target facility in the regional map model; Determine the length of the walking path to determine the walkable distance between the location of the listed store and any target facility; The specific method of calculating the store characteristic parameters of each candidate product on the shelves by the parameter calculation module also includes: Determine the surrounding traffic information of the listed store; Determine the surrounding crowd flow portrait corresponding to the candidate listing product according to the surrounding crowd flow information and a preset crowd portrait matching database; Determine the surrounding crowd flow portrait as the surrounding crowd flow feature of the store of the candidate product; and / or, Determine the surrounding traffic information of the listed store; Determine a number of target facilities and a number of competing stores corresponding to the listed store; Determine the surrounding traffic information of the target facility and the competing store; Calculate the crowd flow overlap between the surrounding crowd flow information of the listed store and the surrounding crowd flow information of any of the target facilities or the competing stores; Determine the overlap of all the crowd flows as the crowd flow characteristics around the store of the candidate product; The surrounding crowd flow information is the device information or image information of all users within the corresponding crowd flow coverage area, the device information is the device ID, the image information is the user's face image or full-body image, and the crowd portrait matching database includes the corresponding relationship between the device information or image information of multiple users and the user's portrait features, and the user's portrait features include: the user's gender, occupation, hobby, specialty, interest, and age; The method for determining the surrounding pedestrian flow information is: calculating the pedestrian flow coverage of the listed store, the target facility or the competing store, and calculating the device information or image information of all users within the pedestrian flow coverage area in the area corresponding to the pedestrian flow coverage area, wherein the pedestrian flow coverage area is determined according to the passenger flow influence of the listed store, the target facility or the competing store; The crowd flow overlap is: the ratio of the number of identical user device information in the surrounding crowd flow information of the listed store and all user device information in the surrounding crowd flow information of any target facility or competing store to the total number of all user device information in the surrounding crowd flow information of the listed store; or The similarity between all user image information in the surrounding traffic information of the listed store and all user image information in the surrounding traffic information of any target facility or competing store.

6. A device for determining shelf goods, characterized in that: The device comprises: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the method for determining shelf goods as described in any one of claims 1-4.

Images