A store price region division method and system

Abstract

This invention discloses a method and system for dividing store price zones. The method includes: S1: acquiring store geographical location data, filtering out outlier stores, and creating an independent price zone for each outlier store; S2: for the remaining non-outlier stores, determining the price zone quota quantity k for that quota unit based on the average store density within that quota unit; S3: mapping and offsetting the store's geographical coordinates based on the customer group correlation among the non-outlier stores and in conjunction with business constraint rules to obtain the offset coordinates; S4: using the offset coordinates as input, and using the price zone quota quantity k as the cluster number, performing cluster analysis, and using each cluster and the independent price zone together as the final price zone division result. This invention achieves a scientific, reasonable, and business-interpretive price zone division by integrating geographical distribution characteristics, customer group correlation, and business constraint rules.

Description

Technical Field

[0001] This invention relates to the field of information mining technology, and in particular to a method and system for dividing store prices into regions. Background Technology

[0002] Traditionally, the classification of chain stores is relatively simple, mainly based on geographical division. However, for medium and large cities, the division of cities is more complex, with significant differences in population density and consumption patterns. Simply dividing by geography cannot provide good data support for store analysis.

[0003] To further improve the accuracy of store segmentation, a price zoning analysis method was introduced. Price zoning is a fundamental step for retail enterprises to achieve refined price management. Its core objective is to divide price zoning based on the similarity of customer groups, customize inventory for different customer groups, and conduct market research and pricing based on the competitive situation in the price zoning, thereby balancing the effectiveness and workload of market research while maintaining a unified price image.

[0004] In the history of related technological development, price zoning methods have evolved from simple geographical boundary delineation to data-driven intelligent delineation. This evolution has mainly gone through the following three stages: The manual experience-based segmentation stage: In the early stages, the segmentation of store prices by retail enterprises relied entirely on the subjective experience of business personnel. The segmentation was based on intuitive judgments of business districts, store distribution, and local consumption characteristics. This method lacked a unified quantitative standard, the segmentation results were inconsistent, and it was difficult to adapt to the rapid growth in the number of enterprise stores.

[0005] Rule-based geographic segmentation stage: With the popularization of Geographic Information Systems (GIS), the industry has developed rule-based segmentation methods based on geographic features, such as dividing regions according to administrative boundaries, transportation networks, and natural boundaries. This method improves the objectivity of segmentation, but does not consider the density of store locations and the correlation of customer groups. In administrative areas with dense stores, it is easy to disrupt the uniformity of price image, and the acquisition cost of accurate transportation network and natural boundary data is relatively high.

[0006] Data-driven quantitative segmentation stage: The development of big data and machine learning technologies has driven the segmentation of store price regions into the quantitative analysis stage. The mainstream methods are divided into two categories. One is the multi-factor evaluation method that refers to price-related characteristics and the homogeneous area method based on spatial differences. The other is the cluster analysis method that first divides the regions and then describes the price characteristics. The methods in this stage have introduced quantitative indicators and have become the current industry mainstream, but they still have many shortcomings and are far from meeting the actual business needs of enterprises.

[0007] The aforementioned development to data-driven quantitative analysis also has the following drawbacks: 1. Limitations of multi-factor evaluation method and spatial interpolation homogeneous zone method: Multi-factor evaluation method requires the acquisition of key natural and socio-economic factors affecting prices (such as population density, income level, consumption capacity, etc.), which is costly for enterprises to acquire data. Moreover, such macro data is poor in terms of timeliness and granularity, making it difficult to support refined price management. The homogeneous zone method based on spatial interpolation requires known price data before dividing the region, while the actual business needs are to divide the region first and then set prices. The two are contradictory in terms of implementation order and cannot meet the actual business process.

[0008] 2. Deficiencies of traditional clustering analysis methods: In clustering algorithms such as K-Means, the hyperparameter k is usually determined by statistical indicators such as elbow plots or silhouette coefficients, which lack business interpretability; if single-objective clustering is performed based solely on the latitude and longitude of stores, insufficient consideration of customer group similarity may result in geographically close stores with large differences in customer groups being grouped into the same price area; if multi-objective optimization is adopted to consider both geographical distance and customer group similarity, the geographical clustering effect will be greatly degraded, which is not conducive to maintaining the consistency of price image.

[0009] 3. Difficulties in the practice of rule-based division: When using administrative boundaries for division, in densely populated areas, the density of stores within the same administrative region may be too high, forcing them to be divided into the same price zone, which undermines the uniformity of price image; when using a combination of transportation network and natural boundaries for division, it is difficult for enterprises to obtain accurate data on road network and natural barriers, and the selection criteria for roads and natural barriers as dividing lines are difficult to formulate, lacking operability.

[0010] 4. Improper handling of outliers: Existing methods usually include outliers (stores whose geographical location is significantly far away from other stores) directly in the clustering process. This causes outliers to occupy a limited price range quota, forcing other relatively scattered stores to be merged into the same price range. At the same time, market research results applicable to outliers are often not applicable to stores in densely populated areas, and vice versa, resulting in a lack of universality of market research results within the price range.

[0011] 5. Insufficient adaptability to differences in store density: Traditional clustering methods rely on relative distance rather than absolute distance, which means that cities with different store densities (such as cities with average store spacing of several thousand meters and cities with average store spacing of several hundred meters) may receive the same amount of price zone quotas. This contradicts business understanding (stores within a radius of several thousand meters in business districts, towns, etc. should usually be classified into the same price zone) and cannot adapt to changes in the density of the company's store layout. Summary of the Invention

[0012] The technical solution of this invention is a method for dividing store price zones, comprising the following steps: S1: Obtain the geographical location data of all stores in the target area, filter out outlier stores based on the distance distribution between each store and its nearest store, and create an independent price zone for each outlier store; S2: For the remaining non-outlier stores, determine the price zone quota quantity k of the quota unit based on the average store density within the quota unit; S3: Based on the customer group correlation between each non-out-of-group store and combined with business constraint rules, the geographical coordinates of the stores are mapped and offset to obtain the offset coordinates. S4: Using the offset coordinates as input, and the price area quota quantity k as the cluster number, perform cluster analysis on the non-outlier stores within each quota unit, and use each cluster obtained from the cluster analysis and the independent price area together as the final price area division result.

[0013] Furthermore, in step S1, the specific categories of outlier stores identified include: S11: Calculate the distance between each store within the enterprise and the nearest store. ; S12: Calculate all distances mean and standard deviation ; S13: Will satisfy > + or < - The stores are set as outlier stores; Furthermore, in step S2, the method for determining the price zone quota quantity k includes: S21: Set the quota unit, where the number of non-outlier stores in the unit is n; S22: For each non-outsider store i, calculate the density_i of the enterprise stores within its business format radiation radius r, where density_i is the number of other enterprise stores within the radiation radius r plus the store i itself; S23: Calculate the average store density for this quota unit. Its formula is:

[0014] S24: Calculate the theoretical price zone quota quantity k using the following formula:

[0015] S25: Round k up and down to obtain the candidate quotas. and .

[0016] Furthermore, the price zone quota quantity k also includes an optimization method, the steps of which are: S241: respectively with and As the number of clusters, K-Means clustering is performed on the offset coordinates of non-outlier stores within the current quota unit, resulting in two clustering results; S242: respectively with and The contour coefficient S is calculated as a hyperparameter, and the contour coefficient is obtained. and contour coefficient ; S243: Select the number of clusters corresponding to the higher silhouette coefficient as the final price zone quota for this quota unit. .

[0017] Furthermore, the method for calculating the profile coefficient S is as follows: S2421: Calculate the minimum average Euclidean distance from store i to any other cluster, using the following formula:

[0018] S2422: Calculate the average Euclidean distance from store i to all other stores in its cluster using the following formula:

[0019] S2423: Calculate each point in the target cluster sample The corresponding profile coefficient is calculated using the following formula:

[0020] S2424: Calculate the mean of the profile coefficient s_i for each store i, and obtain the profile coefficient. .

[0021] Furthermore, in step S3, customer group relevance is quantified using a member overlap rate vector, and the steps include: S301: Quantify the member overlap rate between stores into a vector. , Let be the vector of member overlap rates between the i-th store and the j-th store; S302: Calculate the magnitude of the vector. The calculation formula is:

[0022] in, For the first Membership of each store For the first Membership of each store The total number of members in both stores; S303: The first The store and the first The distance between stores is represented as a polar coordinate vector. , | | represents the actual distance between stores i and j. Let i be the azimuth angle between stores i and j.

[0023] Furthermore, in step S3, the method for mapping and offsetting the geographical coordinates of the store includes: S311: Calculate the offset o_ij of store j relative to store i, using the following formula:

[0024] in This is a dynamic correction function for the offset. Let i be the walking time between stores i and j, and azimuth angle after offset ; S312: Calculate the total offset of store i Its formula is:

[0025] Will Decomposed into longitude components and latitude components ; S313: Calculate the latitude and longitude coordinates of store i after offset. Its formula is:

[0026] in Here are the original latitude and longitude coordinates of store i.

[0027] Furthermore, the dynamic correction function The Ebbinghaus forgetting curve is denoted by the following formula:

[0028] in, , These are the maximum and minimum profit / loss coefficients for member overlap rates, respectively. , , , , It is an adjustable parameter. , , This represents the walking time threshold.

[0029] This invention also provides a store price region division system to implement the store price region division method, including a data acquisition unit, a quota calculation unit, a coordinate mapping unit, and a clustering division unit, wherein: The data acquisition unit is used to acquire the geographical location data of all stores in the target area, filter out outlier stores based on the distance distribution between each store and its nearest store, and create an independent price zone for each outlier store. The quota calculation unit is used to determine the price zone quota quantity k of the remaining non-outlier stores based on the average store density within the quota unit. The coordinate mapping unit is used to map and offset the geographical coordinates of stores based on the customer group correlation between each non-outsider store and in combination with business constraint rules, so as to obtain the offset coordinates. The clustering unit is used to take the offset coordinates as input, use the price area quota quantity k as the number of clusters, perform cluster analysis on the non-outlier stores within each quota unit, and take each cluster obtained from the cluster analysis and the independent price area together as the final price area division result.

[0030] This application also provides an electronic device, including: a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the store price zoning method as described above.

[0031] In practical applications, the modules described in the methods and systems disclosed in this invention can be deployed on a single target server, or each module can be deployed independently on different target servers. In particular, as needed, to provide more powerful computing capabilities, the modules can also be deployed on a cluster of target servers.

[0032] Therefore, this invention achieves a scientific, reasonable, and business-interpretable price zoning by integrating geographical distribution characteristics, customer group correlations, and business constraint rules.

[0033] To provide a clearer and more comprehensive understanding of the present invention, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0035] Figure 1 This is a schematic diagram of the store price zone division method according to an embodiment of the present invention. Detailed Implementation

[0036] Existing technological limitations make it difficult for current price zoning methods to simultaneously ensure price consistency, customer similarity, and the applicability of market research results, thus failing to meet the actual needs of retail enterprises for refined price management.

[0037] This invention does not simply group stores geographically, but integrates geographical distribution characteristics and uses customer group correlation (member overlap rate) and business constraints (such as walking time) to make the clustering results more consistent with the "price influence range" in business scenarios. It maps the geographical distribution of stores to price management areas, thereby effectively improving the accuracy of store segmentation.

[0038] Please see Figure 1 To address the shortcomings of existing technologies, this invention proposes a method for dividing store price zones, comprising the following steps: S1: Obtain the geographical location data of all stores in the target area, filter out outlier stores based on the distance distribution between each store and its nearest store, and create an independent price zone for each outlier store; S2: For the remaining non-outlier stores, determine the price zone quota quantity k of the quota unit based on the average store density within the quota unit; S3: Based on the customer group correlation between each non-out-of-group store and combined with business constraint rules, the geographical coordinates of the stores are mapped and offset to obtain the offset coordinates. S4: Using the offset coordinates as input, and the price area quota quantity k as the cluster number, perform cluster analysis on the non-outlier stores within each quota unit, and use each cluster obtained from the cluster analysis and the independent price area together as the final price area division result.

[0039] The technical solution of this application will be described below with reference to various preferred embodiments and implementation methods.

[0040] In this embodiment, the target area is a retail pharmacy in a certain city. The quota unit is set to the city. The business radiation radius r is set to 500 meters based on the industry characteristics of pharmacies. The walking time threshold is set to t0=10min, t1=20min, and t2=30min based on the price comparison behavior characteristics of pharmacy customers. The constants and adjustable parameters of the dynamic correction function are determined through industry research: cmax=1, cmin=0, α0=−0.05, β0=1.5, α1=0.001, β1=20, and λ=2.

[0041] S1: Obtain the geographical location data of all stores within the target area, filter out outlier stores based on the distance distribution between each store and its nearest store, and create an independent price zone for each outlier store.

[0042] In step S1, outlier stores are quantitatively screened based on the distance distribution between stores and their nearest stores using the mean-standard deviation method. An independent price range is created for each outlier store to prevent them from affecting subsequent clustering results. Specifically, the outlier stores selected include: S11: Calculate the distance between each store within the enterprise and the nearest store. .

[0043] Obtain the latitude and longitude geographic location data of all pharmacies in the city, calculate the straight-line distance from each pharmacy i to other pharmacies within the enterprise based on a Geographic Information System (GIS), and select the minimum value as the distance from each pharmacy i to the other pharmacies within the enterprise. (Unit: meter) Let i be the distance between store i and the nearest store.

[0044] S12: Calculate the mean distance and standard deviation : Collect all stores Forming distance datasets Calculate the mean of the dataset. and standard deviation , For example, calculated rice, rice.

[0045] S13: Identify and filter outoutstanding stores, and create independent price zones: Will satisfy (for example (meters) or (for example The store with a mileage of 1000 meters was identified as an outlier store.

[0046] In this embodiment, a suburban store Meters (greater than 280 meters), a store in a core business district Stores smaller than 120 meters were identified as outliers. A separate price zone was created for each outlier store, and this zone was designated as "one store, one zone" and would not be included in subsequent cluster analysis.

[0047] This step uses the mean and standard deviation method to quantitatively screen outlier stores. The rules are clear and controllable, which is in line with the business understanding of retail enterprises and avoids the complex parameter tuning of noisy clustering algorithms such as HDBSCAN. Creating independent price zones for outlier stores frees up clustering quotas, preventing outlier stores from occupying quotas and forcing other stores into clustering. Outlier stores are surveyed independently, avoiding the predicament of applying market research results from densely populated areas to outlier stores, thus laying a high-quality data foundation for the accuracy of subsequent clustering analysis.

[0048] S2: For the remaining non-outlier stores, determine the price zone quota quantity k of the quota unit based on the average store density within the quota unit.

[0049] This step calculates the theoretical price zone quota k based on the average store density of non-outlier stores within the quota unit, and obtains the final quota through contour coefficient optimization. To ensure that the k-value is adapted to the store density and has business interpretability, the specific steps include S21-S25: S21: Set the quota unit, where the number of non-outlier stores in the unit is n.

[0050] The quota unit can be selected from administrative units such as provinces and cities, or internal operating units of enterprises. Let n be the number of non-outlier stores within the current quota unit.

[0051] S22: Calculate the radiation density of a single store.

[0052] For each non-outlier store i, calculate the number of business stores within its 500-meter radius (including stores). (Itself), that is, the store's For example, if store A in a prime commercial area has 10 other stores within 500 meters, then... For suburban store B, if there are two other businesses' stores within 500 meters, then... .

[0053] S23: Calculate the average store density.

[0054] Calculate the average store density for this quota unit. Its formula is:

[0055] In the formula, This refers to the number of non-outsider stores within the quota unit. Let be the radiation density of the i-th store.

[0056] S24: Calculate the theoretical price zone quota quantity k using the following formula:

[0057] According to the formula, when the store density is high, the k value is small and the number of price zones is small, which is conducive to maintaining a consistent price image. When the store density is low, the k value is large and the number of price zones is large, which is conducive to refined management.

[0058] Furthermore, this application further optimizes the value of k, and the optimization method includes: S241: respectively with and To determine the number of clusters, K-Means clustering was performed on the offset coordinates of non-outlier stores, resulting in two clustering results; in this embodiment... Therefore, only one clustering result is obtained; S242: Calculate the silhouette coefficients of the two clustering results. and The method for calculating the contour coefficient is explained in detail later in this embodiment; The silhouette coefficient is a core indicator for measuring clustering performance. In the embodiments of this application, the silhouette coefficient... and The calculation method uses the following steps S2421-S2424: S2421: Calculate the minimum average Euclidean distance from store i to any other cluster, using the following formula:

[0059] The higher the bi value, the lower the similarity between the stores and other clusters, and the better the separation.

[0060] S2422: Calculate the average Euclidean distance from store i to all other stores in its cluster using the following formula:

[0061] Reflects the cohesion of the cluster. The smaller the value, the higher the similarity between the store and other stores in the cluster, and the better the cohesion.

[0062] S2423: Calculate each point in the target cluster sample The corresponding profile coefficient is calculated using the following formula:

[0063] The value range is [−1, 1]. The closer the value is to 1, the better the clustering effect of the stores.

[0064] S2424: Calculate the mean of the profile coefficient s_i for each store i, and obtain the profile coefficient. , The closer the result is to 1, the better the cohesion and separation of the overall clustering results.

[0065] S243: Select the number of clusters with higher silhouette coefficients as... In this embodiment .

[0066] S25: Round k up and down to obtain the candidate quotas. and For example, if k is an integer 20, then , If k is not an integer (e.g., k=20.6), then , .

[0067] Therefore, this application calculates the k-value based on store density, giving the k-value clear business interpretability. This solves the problem of traditional methods selecting k-values ​​using elbow plots, which ignores store density and contradicts business understanding. Furthermore, optimizing the k-value through the silhouette coefficient ensures the cohesion and separation of the clustering results. The k-value can adapt to the development of the enterprise's store layout: when store density increases, the k-value decreases, which is beneficial for price and image consistency; when store density decreases, the k-value increases, which is beneficial for refined management.

[0068] S3: Based on the customer group correlation between each non-out-of-group store and combined with business constraint rules, the geographical coordinates of the stores are mapped and offset to obtain the offset coordinates.

[0069] This step quantifies the customer group correlation between stores by measuring the member overlap rate. Combined with business constraint rules (a dynamic correction function based on walking time, such as a walking time of no more than 30 minutes), the original latitude and longitude of the stores are mapped and offset, so that stores with similar customer groups are more concentrated in the mapping space, while avoiding unreasonable offsets.

[0070] Customer group relevance is quantified using a member overlap rate vector, and the steps include: S301: Select member consumption data from non-outlier stores in the city, and quantify the member overlap rate between stores (between store i and store j) into a vector. . The direction is consistent with the direction from store i to store j, and the modulus reflects the degree of overlap of members.

[0071] S302: Calculate the magnitude of the vector. The calculation formula is:

[0072] in, For the first Membership of each store For the first Membership of each store Let $\frac{1}{j}$ be the number of members shared by both stores. For example, if store i has 1000 members and store j has 500 members shared by both stores, then $\frac{1}{j}$ is the number of members shared by both stores. =500 / 1000=0.5.

[0073] S303: The first The store and the first The distance between stores is represented as a polar coordinate vector.

[0074] | | represents the actual distance between stores i and j. Let i be the azimuth angle between stores i and j.

[0075] In this step, the methods for mapping and offsetting the geographical coordinates of the stores include: S311: Calculate the offset o_ij of store j relative to store i, using the following formula:

[0076] in This is a dynamic correction function for the offset. Let i be the walking time between stores i and j, and azimuth angle after offset ; In this embodiment, the dynamic correction function The Ebbinghaus forgetting curve is used to adjust the membership overlap rate based on walking time, aligning with the behavior of pharmacy customers comparing prices within a 30-minute walk. Its formula is:

[0077] in, , , where are constants, and represent the maximum and minimum profit / loss coefficients (also called influence coefficients or correction coefficients) for member overlap rates, respectively. , , , , These are adjustable parameters, determined through experiments or research based on actual business needs. , , This represents the walking time threshold.

[0078] The specific meaning and example of this formula are as follows: when hour, Member overlap rate results in no discounts, and customer group relevance has the greatest impact. when hour, Member overlap rate decreases linearly; when hour, The membership overlap rate decays rapidly and non-linearly. when hour, The overlap rate of members was completely eliminated, but the relevance of customer groups was not affected.

[0079] Calculation rules: A scalar vector that overlaps with the member vector. This is a scalar multiplication, and the result is an offset vector. .

[0080] Constraints: The offset angle This ensures that the store's offset direction is consistent with its original position, avoiding reverse offset.

[0081] S312: Calculate the total offset of store i Its formula is: , Will Decomposed into longitude components and latitude components These figures reflect the store's offset in the longitude and latitude directions, respectively.

[0082] S313: Calculate the latitude and longitude coordinates of store i after offset. Its formula is:

[0083] in The original latitude and longitude coordinates of store i, and the offset latitude and longitude coordinates. This serves as input for subsequent cluster analysis.

[0084] This step quantifies customer group correlation by using member overlap rate vectors, thereby achieving a quantitative expression of customer group characteristics and making it possible to calculate the comprehensive customer group impact of multiple stores on a single store.

[0085] Furthermore, by using a dynamic correction function, combined with the walking price comparison behavior characteristics of pharmacy customers, and introducing business constraints, unreasonable situations such as excessive offset and reverse offset are avoided, making the coordinate offset more in line with the business scenario.

[0086] Furthermore, after the coordinate shift, stores with similar customer groups are more concentrated in the mapping space. Subsequent single-objective clustering can take into account both geographical and customer group characteristics, avoiding the degradation of clustering effect caused by multi-objective optimization and greatly improving computational efficiency.

[0087] S4: Using the offset coordinates as input and the price zone quota quantity k as the cluster number, perform cluster analysis on the non-outlier stores within each quota unit. Combine each cluster obtained from the cluster analysis with the independent price zone to form the final price zone division result. The specific implementation is as follows: Selecting initial cluster centers: coordinates after offset from non-outlier stores. Random selection For example, use the coordinates as the initial cluster centers. Then, 20 initial cluster centers are selected.

[0088] Stores are assigned to clusters: the offset coordinates of each non-outlier store are calculated to the Euclidean distance from each initial cluster center, and the store is assigned to the nearest cluster.

[0089] Update cluster centers: Calculate the mean longitude and mean latitude of the offset coordinates of all stores in each cluster, and use the mean as the new cluster center.

[0090] Iterative convergence: Repeat the above steps of "store allocation" and "update cluster center" until the change in the position of the cluster center is less than the preset convergence threshold (set to 0.001 degrees in this embodiment), then stop the iteration and obtain the final cluster.

[0091] Generate the final division result: Each cluster obtained by K-Means clustering is taken as a price region and combined with the independent price regions of outlier stores in step S1 to obtain the final store price region division result for the city.

[0092] This step combines clustering with independent price zones to achieve a final division result that is suitable for managing multiple stores in one zone in densely populated areas, as well as managing one store in one zone for out-of-area stores, thus balancing the needs of unified corporate pricing and image with refined management.

[0093] After completing the price region division, the company selects market research targets in each price region according to its internal benchmarking criteria for competitors and benchmark products. After obtaining the market research target data, the company adjusts the product prices based on the product type and competition index.

[0094] Therefore, the technical effects achieved by this application are as follows: 1. It integrates the consideration of multiple dimensions such as geographical distribution, store density, and customer group correlation, and solves the technical defects of traditional segmentation methods that ignore customer groups in single-objective optimization and lack business interpretability of cluster hyperparameters; 2. By using a dynamic correction function based on walking time to impose business constraints on coordinate offsets, unreasonable situations such as excessive offsets and reverse offsets are avoided, making the coordinate mapping more in line with actual consumption scenarios. At the same time, single-objective clustering can take into account multi-dimensional needs, avoiding the degradation of clustering results caused by multi-objective optimization and significantly improving computational efficiency. 3. Outlier store screening and cluster quota optimization were achieved by using the mean-standard deviation method and the silhouette coefficient, respectively, which ensured the quantitative controllability of the division process and the cohesion and separation of the cluster results, thereby improving the scientificity and rationality of the regional division. Based on the above embodiments, the present invention also provides a store price region division system, including a data acquisition unit, a quota calculation unit, a coordinate mapping unit, and a clustering division unit, wherein: The data acquisition unit is used to acquire the geographical location data of all stores in the target area, filter out outlier stores based on the distance distribution between each store and its nearest store, and create an independent price zone for each outlier store. The quota calculation unit is used to determine the price zone quota quantity k of the remaining non-outlier stores based on the average store density within the quota unit. The coordinate mapping unit is used to map and offset the geographical coordinates of stores based on the customer group correlation between each non-outsider store and in combination with business constraint rules, so as to obtain the offset coordinates. The clustering unit is used to take the offset coordinates as input, use the price area quota quantity k as the number of clusters, perform cluster analysis on the non-outlier stores within each quota unit, and take each cluster obtained from the cluster analysis and the independent price area together as the final price area division result.

[0095] This invention also provides an electronic device, including: a processor, a storage medium, and a bus. The storage medium stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the store price zone division method as described above.

[0096] The technical solution of this application can adapt to the development and changes in the enterprise's store layout, effectively balancing the needs of unified pricing image and refined price management; it achieves segmentation based on the enterprise's own data, reducing data acquisition costs, while improving the applicability of market research results and reducing the workload of market research; the algorithm has excellent portability and can be adapted to the store segmentation needs of various chain retail formats such as retail pharmacies, supermarkets, and convenience stores; the system and equipment are highly deployable, reducing the enterprise's system transformation costs, and providing a scientific basis for regional segmentation for the enterprise's pricing, pallet customization and other business operations.

[0097] It should be noted that those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, which may include, but is not limited to, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0098] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for dividing store prices into regions, characterized in that, Includes the following steps: S1: Obtain the geographical location data of all stores in the target area, filter out outlier stores based on the distance distribution between each store and its nearest store, and create an independent price zone for each outlier store; S2: For the remaining non-outlier stores, determine the price zone quota quantity k of the quota unit based on the average store density within the quota unit; S3: Based on the customer group correlation between each non-out-of-group store and combined with business constraint rules, the geographical coordinates of the stores are mapped and offset to obtain the offset coordinates. S4: Using the offset coordinates as input, and the price area quota quantity k as the cluster number, perform cluster analysis on the non-outlier stores within each quota unit, and use each cluster obtained from the cluster analysis and the independent price area together as the final price area division result.

2. The method for dividing store price zones as described in claim 1, characterized in that, step In S1, outlier stores are specifically identified as follows: S11: Calculate the distance between each store within the enterprise and the nearest store. ; S12: Calculate all distances mean and standard deviation ; S13: Will satisfy > + or < - The stores are set as outlier stores.

3. The method for dividing store price zones as described in claim 2, characterized in that: In step S2, the method for determining the price zone quota quantity k includes: S21: Set the quota unit, where the number of non-outlier stores in the unit is n; S22: For each non-outsider store i, calculate the density_i of the enterprise stores within its business format radiation radius r, where density_i is the number of other enterprise stores within the radiation radius r plus the store i itself; S23: Calculate the average store density for this quota unit. Its formula is: ； S24: Calculate the theoretical price zone quota quantity k using the following formula: ； S25: Round k up and down to obtain the candidate quotas. and .

4. The method for dividing store price zones as described in claim 3, characterized in that, The price zone quota quantity k also includes an optimization method, the steps of which are: S241: respectively with and As the number of clusters, K-Means clustering is performed on the offset coordinates of non-outlier stores within the current quota unit, resulting in two clustering results; S242: respectively with and The contour coefficient S is calculated as a hyperparameter, and the contour coefficient is obtained. and contour coefficient ; S243: Select the number of clusters corresponding to the higher silhouette coefficient as the final price zone quota for this quota unit. .

5. The method for dividing store price zones as described in claim 4, characterized in that, The method for calculating the profile coefficient S is as follows: S2421: Calculate the minimum average Euclidean distance from store i to any other cluster, using the following formula: ； S2422: Calculate the average Euclidean distance from store i to all other stores in its cluster using the following formula: ； S2423: Calculate each point in the target cluster sample The corresponding profile coefficient is calculated using the following formula: ； S2424: Calculate the mean of the profile coefficient s_i for each store i, and obtain the profile coefficient. .

6. The method for dividing store price zones as described in claim 1, characterized in that, In step S3, customer group relevance is quantified using a member overlap rate vector, and the steps include: S301: Quantify the member overlap rate between stores into a vector. , Let be the vector of member overlap rates between the i-th store and the j-th store; S302: Calculate the magnitude of the vector. The calculation formula is: ； in, For the first Membership of each store For the first Membership of each store The total number of members in both stores; S303: The first The store and the first The distance between stores is represented as a polar coordinate vector. , | | represents the actual distance between stores i and j. Let i be the azimuth angle between stores i and j.

7. The method for dividing store price zones as described in claim 1, characterized in that, In step S3, the method for mapping and offsetting the geographical coordinates of the store includes: S311: Calculate the offset o_ij of store j relative to store i, using the following formula: ； in This is a dynamic correction function for the offset. Let be the walking time between stores i and j, and azimuth angle after offset ; S312: Calculate the total offset of store i Its formula is: ； Will Decomposed into longitude components and latitude components ; S313: Calculate the latitude and longitude coordinates of store i after offset. Its formula is: ； in Here are the original latitude and longitude coordinates of store i.

8. The method for dividing store price zones as described in claim 7, characterized in that, The dynamic correction function The Ebbinghaus forgetting curve is denoted by the following formula: ； in, , These are the maximum and minimum profit / loss coefficients for member overlap rates, respectively. , , , , It is an adjustable parameter. , , This represents the walking time threshold.

9. A store price zoning system for implementing the store price zoning method according to any one of claims 1-8, characterized in that, It includes a data acquisition unit, a quota calculation unit, a coordinate mapping unit, and a clustering unit, wherein: The data acquisition unit is used to acquire the geographical location data of all stores in the target area, filter out outlier stores based on the distance distribution between each store and its nearest store, and create an independent price zone for each outlier store. The quota calculation unit is used to determine the price zone quota quantity k of the remaining non-outlier stores based on the average store density within the quota unit. The coordinate mapping unit is used to map and offset the geographical coordinates of stores based on the customer group correlation between each non-outsider store and in combination with business constraint rules, so as to obtain the offset coordinates. The clustering unit is used to take the offset coordinates as input, use the price area quota quantity k as the number of clusters, perform cluster analysis on the non-outlier stores within each quota unit, and take each cluster obtained from the cluster analysis and the independent price area together as the final price area division result.

10. An electronic device, comprising: The device includes a processor, a storage medium, and a bus. The storage medium stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the store price zoning method as described above.

Images