Store evaluation method based on indices related to the surrounding environment of the store.
The method quantifies geographic themes using statistical modeling and linear regression to create an SSTI, addressing the inefficiencies and subjectivity of conventional store evaluations, offering a cost-effective and objective assessment of store environments.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- HANGZHOU HUMPBACK WHALE TECHNOLOGY CO LTD
- Filing Date
- 2024-03-26
- Publication Date
- 2026-04-10
AI Technical Summary
Conventional store evaluation methods are costly and subjective, relying on labor-intensive field surveys and expert evaluations that lack completeness and objectivity, making them unsuitable for large-scale assessments.
A method that quantifies geographic environment themes using a combination of geographic information and natural features, employing statistical modeling and linear regression to generate a large-scale spatiotemporal index (SSTI) for evaluating store management status, which includes collecting data, statistically modeling, extracting theme characteristic indices, and constructing a linear regression model for real-world applications.
Provides a scientific, efficient, and objective evaluation of store environments, reducing labor costs, eliminating subjective influences, and improving the accuracy and speed of store assessments by using interpretable geographic indicators.
Smart Images
Figure 2026511188000001_ABST
Abstract
Description
[Technical Field]
[0001] This invention relates to the technical field of store commercial evaluation, and more specifically to a store evaluation method using indices related to the surrounding environment of a store. [Background technology]
[0002] Commercial users, when performing actions such as selecting store addresses, grouping stores by image, and conducting offline store comparison tests, typically need to have an objective understanding of the surrounding geographical environment, commercial environment, and people to facilitate the evaluation process. The large-scale spatiotemporal store index is a set of related indices that evaluate the convenience and comfort of people's outings, along with an abstracted large-scale geographical environment theme score for stores with a certain degree of business guidance significance, derived from statistical modeling of surrounding geographical information and weather data. It perfectly meets the needs of commercial users, provides an objective and clear understanding of the surrounding environment, and is advantageous in improving the efficiency of the evaluation process. Most conventional evaluation methods involve collecting geographical information through field surveys by workers and then obtaining scores for the geographical environment of a region through expert evaluation voting. However, field information collection by workers incurs very high labor and time costs, and the completeness of the information cannot be guaranteed, making it unsuitable as an evaluation method for large areas. Furthermore, geographical environment scores obtained through expert evaluation are highly subjective, and the evaluation results are heavily influenced by individual factors. [Overview of the project] [Problems that the invention aims to solve]
[0003] This invention provides a method for evaluating stores using relevant indices for the surrounding environment. It effectively solves the problems in the aforementioned background technology, where on-site information collection by workers is extremely costly in terms of labor and time, and the completeness of information collection cannot be guaranteed, making it unsuitable as an evaluation method for a wide area, and where geographical environment scores obtained by experts are too subjective, causing evaluation results to be heavily influenced by individual factors. [Means for solving the problem]
[0004] To achieve the above objective, the present invention provides the following technical solution: A method for evaluating stores using an index related to the surrounding environment of a store, which quantifies geographic environment themes that have business value and are highly interpretable, mainly from geographic information, and uses the combination of geographic information and related natural features as an index score to evaluate the store's management status or capacity to meet management needs, thereby providing a basis for reference in the store business scene. Step S1 involves collecting geographic information data around offline stores, Step S2 involves statistically modeling the entities surrounding the store, Step S3 involves rationally extracting the theme characteristic index SSTI, Step S4 involves constructing a linear regression model evaluation system, This includes step S5, which involves applying the SSTI metric to real-world business scenarios.
[0005] According to the above proposed technology, in S1, geographic information data around the offline store is obtained by mainly using different data sources during collection, thereby obtaining geographic information around the target store, mainly including information about the surrounding environment and related natural features. After acquiring geographical information, facilities near the target store are divided into eight categories: transportation, housing, shopping, dining, travel, lifestyle, work, and entertainment. For each category, the number of facilities related to that category near the target store and the distance from the target store are obtained.
[0006] According to the above proposed technology, in S2, when statistically modeling the entities surrounding the store, a theme intensity model and a theme coherence model are mainly established, and by combining the results of the two models, a large-scale spatiotemporal index under the theme is generated. Using brand data as an example, we will verify the magnitude of the impact of large-scale spatiotemporal indicators on customer numbers using a linear regression model. If the verification is successful, large-scale spatiotemporal indicators can be applied to the real-world business scenarios of brands, contributing to improved efficiency in brands' evaluation of the surrounding environment of stores, and ultimately serving as an important reference for demand such as store address selection.
[0007] According to the above technical solution, in S2, specifically when performing statistical modeling, the store geographical location environmental theme score is affected by the number intensity and concentration of theme interest points around the store. Taking the POIs around the store as samples, the following model is established for the distribution of theme POIs around the store:
Number
Number
Number
Number
[0008] According to the above proposed technology, in S2, the model-based calculation process includes input, output, and calculation steps, and specifically, is as follows: Input:t:Evaluation themes including a total of 8 categories: transportation, housing, shopping, dining, travel, lifestyle, work, and entertainment. m: Total number of stores to be evaluated. n t (s) :t The number of POIs queried in s stores under the theme, N t (s) :t The number of POIs queried in s stores under the theme, p t (s) Initial probability parameters of the POI intensity model for s stores under the theme: σ t (s) : The standard deviation of all POIs and store distances for s stores under theme t. (x i (s) ,y i (s) ):s is the coordinate of the iPOI point near the store relative to the store, α and β are the inter-store strength model priors, following a beta distribution, and are model variables. μ and λ are prior variables of the inter-store settling ratio model, following a Gaussian distribution. Output: SSTI t (s) :s stores' scores under the theme t.
[0009] According to the above technical proposal, the calculation process in S2 is as follows: 1) Execute at each t, 2) Start the loop, 3) In the intensity model, the intensity of themes and points of interest (POIs) around stores follows a Bernoulli distribution, and its probability parameter follows a beta distribution. The integrated probability density function of the intensity model is as follows:
number
number
number
number
[0010] According to the above proposed technology, in S3, the rational extraction of the theme characteristic index SSTI is mainly done by using maximum likelihood estimation for each store and each theme.
number
number
number
number
number
number
[0011] According to the above proposed technology, in S4, the linear regression model evaluation system specifically selects indicators that can be used for monitoring and that can evaluate the store management status or the target of management, such as the number of customers and sales volume, and then establishes a model using geographic information indicators and other relevant natural features, and embodies the rationality of the geographic information indicators through the model fitting degree gain. Linear regression models offer good linear interpretability and feature compatibility, possess a complete quantified feature evaluation and inference mechanism, can easily describe the gain of feature superposition, have a natural fit for evaluating large-scale spatiotemporal features of offline stores, and do not require preconditions for training and test sets. Therefore, linear regression models evaluate the magnitude and positive / negative influence of independent variable features on the dependent variable.
[0012] According to the above technical proposal, in S4, the coefficient of determination is used as an index to evaluate the degree of model fitting, representing the percentage that can be interpreted as a regression relationship in the total change of the target variable. The closer the value is to 1, the better the fitting effect. The Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) can evaluate the complexity of the model being estimated and the quality of the model fitting data. The smaller the value, the better the model effect. Using store information, real-time weather, and eight types of geographic environmental indicators as independent variables, and customer count data as the dependent variable, a linear regression model was constructed. After adding each of the eight geographic environmental features to the evaluation model, R 2 The effect of the traffic index was most evident, followed by the entertainment index and the travel index. Interpretation of the target variable—changes in the number of visitors—was enhanced, and both the AIC and BIC for each indicator decreased, indicating good model fitting and no overfitting.
[0013] According to the above proposed technology, in S5, the SSTI index is applied to a real business scenario, and based primarily on the above verification results, the brand is provided with a certain degree of business guidance and reference value regarding store address selection and store grouping. When other similar conditions exist, new store addresses should prioritize locations with convenient transportation access and some entertainment facilities. For example, commercial complexes and tourist destinations may also be considered. Depending on the needs of the real-world scenario, the store's spatiotemporal scale indicator provides objective numerical evidence for commercial users to quantify the surrounding environment and quickly improves the efficiency of geographical evaluation. [Effects of the Invention]
[0014] Beneficial advantages of the present invention compared to conventional technology: The present invention has a scientific and rational structure, and is safe and convenient to use.
[0015] 1. By quantifying the geographical information and natural environment surrounding a target area into numerical values encompassing eight dimensions—transportation, housing, shopping, dining, travel, lifestyle, work, and entertainment—users can quickly and efficiently grasp the surrounding environmental conditions. The combination of geographical information and related natural features is used as an indicator score to evaluate the store's operating status or capacity to meet business needs, providing a basis for realistic scenarios such as store address selection, store grouping, and store management. Furthermore, the eight indicators can be used as predictive features of store operating conditions, and regression models can be used to investigate whether the surrounding environment significantly influences key indicators such as customer numbers, and the magnitude of that influence, providing valuable guidance for businesses.
[0016] Furthermore, by using the efficient and stable method of statistical modeling, labor costs can be significantly reduced, the influence of human subjective elements can be eliminated, and large-scale spatiotemporal indicators and expert review are not in conflict. Large-scale spatiotemporal indicators can be used to support experts in making more scientific and objective evaluations, making it easier to obtain better results and enabling the quantification of geo-environment themes that have business value and are highly interpretable from geographic information.
[0017] 2. Linear regression models possess good linear interpretability and feature compatibility, have a complete quantified feature evaluation and inference mechanism, can easily describe the gain of feature superposition, have a natural fit for evaluating large-scale spatiotemporal features of offline stores, do not require preconditions for training sets and test sets, and can more quickly improve the efficiency of geographic evaluation work.
[0018] The drawings are provided to further understand the present invention and, together with the embodiments of the present invention, constitute part of the specification and do not constitute a limitation of the present invention. [Brief explanation of the drawing]
[0019] [Figure 1] This is a step flowchart of the evaluation method of the present invention. [Modes for carrying out the invention]
[0020] Example: As shown in Figure 1, the present invention provides the following technical solution: A method for evaluating stores using an index related to the surrounding environment of a store, which quantifies geographic environment themes that have business value and are highly interpretable, mainly from geographic information, and uses the combination of geographic information and related natural features as an index score to evaluate the store's management status or capacity to meet management needs, thereby providing a basis for reference in the store's business scene. Step S1 involves collecting geographic information data around offline stores, Step S2 involves statistically modeling the entities surrounding the store, Step S3 involves rationally extracting the theme characteristic index SSTI, Step S4 involves constructing a linear regression model evaluation system, This includes step S5, which involves applying the SSTI metric to real-world business scenarios.
[0021] According to the proposed technology described above, in S1, geographic information data around the offline store is obtained by mainly using different data sources during collection, thereby obtaining geographic information around the target store that mainly includes information about the store's surrounding environment and related natural features. After acquiring geographical information, facilities near the target store are divided into eight categories: transportation, housing, shopping, dining, travel, lifestyle, work, and entertainment. For each category, the number of facilities related to that category near the target store and the distance from the target store are obtained.
[0022] According to the above proposed technology, in S2, when statistically modeling the entities surrounding the store, a theme intensity model and a theme coherence model are mainly established, and by combining the results of the two models, a large-scale spatiotemporal index under the theme is generated. Using brand data as an example, we will verify the magnitude of the impact of large-scale spatiotemporal indicators on customer numbers using a linear regression model. If the verification is successful, large-scale spatiotemporal indicators can be applied to the real-world business scenarios of brands, contributing to improved efficiency in brands' evaluation of the surrounding environment of stores, and ultimately serving as an important reference for demand such as store address selection.
[0023] According to the above technical proposal, when performing statistical modeling in S2, the store's geographical location environment theme score is influenced by the number intensity and degree of aggregation of themes of interest around the store. Using the points of interest around the store as a sample, the following model is established for the distribution of themes of POIs around the store:
number
number
number
number
[0024] According to the above technical solution, in S2, the model-based calculation process includes an input, an output, and a calculation step, specifically as follows: Input: t: An evaluation theme including a total of 8 categories of transportation, housing, shopping, gourmet, travel, life, work, and entertainment, m: The total number of stores to be evaluated, n t (s) : The number of POIs queried at s stores under the t theme, N t (s) : The number of POI queries at s stores under the t theme, p t (s) : The initial probability parameter of the POI intensity model of s stores under the t theme, σ t (s) : The standard deviation of the distance between all POIs and the store at s stores under the t theme, (x i (s) , y i (s) ): The coordinates of the iPOI point near the s store with respect to the store, α, β: The store-intermediate intensity model prior, following a beta distribution, α, β are variables of the model, μ, λ: The store-intermediate aggregation model prior, following a Gaussian distribution, μ, λ are variables of the model, Output: SSTI t (s) : The score of s stores under the t theme.
[0025] According to the above technical solution, the calculation process in S2 is as follows: 1) Execute at each t, 2) Start the loop, 3) In the intensity model, the intensity of themes and points of interest (POIs) around stores follows a Bernoulli distribution, and its probability parameter follows a beta distribution. The integrated probability density function of the intensity model is as follows:
number
number
number
number
[0026] According to the above technical proposal, in S3, the rational extraction of the theme characteristic index SSTI is mainly done by using maximum likelihood estimation for each store and each theme.
number
number
number
number
number
number
[0027] According to the above technical proposal, in S4, the linear regression model evaluation system specifically selects indicators that can be used for monitoring and that can evaluate the store management status or target of management, such as the number of customers and sales volume, and then establishes a model using geographic information indicators and other relevant natural features, and embodies the rationality of the geographic information indicators through the model fitting degree gain. Linear regression models offer good linear interpretability and feature compatibility, possess a complete quantified feature evaluation and inference mechanism, can easily describe the gain of feature superposition, have a natural fit for evaluating large-scale spatiotemporal features of offline stores, and do not require preconditions for training and test sets. Therefore, linear regression models evaluate the magnitude and positive / negative influence of independent variable features on the dependent variable.
[0028] According to the above technical proposal, in S4, the coefficient of determination is used as an index to evaluate the degree of model fitting, representing the percentage that can be interpreted as a regression relationship in the total change of the target variable. The closer the value is to 1, the better the fitting effect. The Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) can evaluate the complexity of the model being estimated and the quality of the model fitting data. The smaller the value, the better the model effect. Using store information, real-time weather, and eight types of geographic environmental indicators as independent variables, and customer count data as the dependent variable, a linear regression model was constructed. After adding each of the eight geographic environmental features to the evaluation model, R 2 The effect of the traffic index was most evident, followed by the entertainment index and the travel index. Interpretation of the target variable—changes in the number of visitors—was enhanced, and both the AIC and BIC for each indicator decreased, indicating good model fitting and no overfitting.
[0029] According to the above technical proposal, in S5, the SSTI metric will be applied to real-world business scenarios, and primarily based on the above verification results, it will provide brands with a certain degree of business guidance and reference value regarding store address selection and store grouping. When other similar conditions exist, new store addresses should prioritize locations with convenient transportation access and some entertainment facilities. For example, commercial complexes and tourist destinations may also be considered. Depending on the needs of the real-world scenario, the store's spatiotemporal scale indicator provides objective numerical evidence for commercial users to quantify the surrounding environment and quickly improves the efficiency of geographical evaluation.
[0030] Finally, although the foregoing is merely a preferred example of the present invention and is not intended to limit it, and despite the detailed description of the present invention with reference to the above embodiments, those skilled in the art can modify the technical proposals described in the above embodiments or replace some of the technical features with equivalents. Any modifications, equivalent substitutions, improvements, etc., made in the spirit and principles of the present invention should be within the scope of protection of the present invention.
[0031] (Note) (Note 1) This method evaluates stores using indices related to the surrounding environment, primarily by quantifying geographical environment themes that have business value and are highly interpretable, based on geographical information. The combination of geographical information and related natural features is used as an indicator score to evaluate the store's management status or capacity to meet management needs, providing a basis for reference in the store's business scenario. Step S1 involves collecting geographic information data around offline stores, Step S2 involves statistically modeling the entities surrounding the store, Step S3 involves rationally extracting the theme characteristic index SSTI, Step S4 involves constructing a linear regression model evaluation system, A method for evaluating stores using an index related to the surrounding environment of a store, characterized by including step S5, which involves applying the index SSTI to a real-world business scenario.
[0032] (Note 2) In S1, the geographic information data around the offline store is collected by mainly using different data sources, thereby obtaining geographic information around the target store, mainly including information about the surrounding environment and related natural features. A method for evaluating stores using an index of the surrounding environment of a store, as described in Appendix 1, characterized by obtaining geographic information, dividing facilities near the target store into eight categories of themes: transportation, residence, shopping, gourmet food, travel, lifestyle, work, and entertainment, and obtaining information on the number of facilities related to that theme near the target store and the distance from the target store for each theme.
[0033] (Note 3) In S2 above, when statistically modeling the entities surrounding the store, a theme intensity model and a theme coherence model are mainly established, and by combining the results of the two models, a large-scale spatiotemporal index under the theme is generated. A method for evaluating stores using indices related to the store's surrounding environment, as described in Appendix 1, characterized by primarily verifying the magnitude of the impact of large-scale spatiotemporal indicators on the number of customers using a linear regression model, and if the verification is successful, applying the large-scale spatiotemporal indicators to the brand's real-world business scenarios, contributing to improving the efficiency of the brand's evaluation of the store's surrounding environment, and ultimately serving as an important reference for demand such as store address selection.
[0034] (Note 4) In S2 above, when performing statistical modeling, the number of store geographical location environment themes is influenced by the number intensity and degree of clustering of themes of interest around the store. Using the points of interest around the store as a sample, the following model is established for the distribution of themes of POIs around the store:
number
number
number
number
[0035] (Note 5) In S2, the model-based computation process includes inputs, outputs, and computation steps, and specifically, it is as follows: Input:t:Evaluation themes including a total of 8 categories: transportation, housing, shopping, dining, travel, lifestyle, work, and entertainment. m: Total number of stores to be evaluated. n t (s) :t The number of POIs queried in s stores under the theme, N t (s) :t The number of POIs queried in s stores under the theme, p t (s) Initial probability parameters of the POI intensity model for s stores under the theme: σ t (s) : The standard deviation of all POIs and store distances for s stores under theme t. (x i (s) ,y i (s) ):s is the coordinate of the iPOI point near the store relative to the store, α and β are the inter-store strength model priors, following a beta distribution, and are model variables. μ and λ are prior variables of the inter-store settling ratio model, following a Gaussian distribution. Output: SSTI t (s) A method for evaluating stores using an index related to the surrounding environment of stores, as described in Appendix 3, characterized in that it is the score of s stores under theme t.
[0036] (Note 6) The calculation process in S2 is as follows: 1) Execute at each t, 2) Start the loop, 3) In the intensity model, the intensity of themes and points of interest (POIs) around stores follows a Bernoulli distribution, and its probability parameter follows a beta distribution. The integrated probability density function of the intensity model is as follows:
number
number
number
number
[0037] (Note 7) In S3, the rational extraction of the theme feature index SSTI is mainly done using maximum likelihood estimation for each store and each theme.
number
number
number
number
number
number
[0038] (Note 8) In S4, the linear regression model evaluation system specifically selects indicators that can be used for monitoring and that can evaluate the store management status or the target of management, then establishes a model using geographic information indicators and other relevant natural features, and embodies the rationality of the geographic information indicators through the model fitting degree gain. The method for evaluating stores using an index of the surrounding environment of a store, as described in Appendix 1, is characterized in that the linear regression model has good linear interpretability and feature compatibility, possesses a complete quantified feature evaluation and inference mechanism, can easily describe the gain of feature superposition, has a natural fit for evaluating large-scale spatiotemporal features of offline stores, and does not require preconditions for training and test sets, so the linear regression model evaluates the magnitude and positive / negative of the influence of independent variable features on the dependent variable.
[0039] (Note 9) In S4, the coefficient of determination is used as an index to evaluate the degree of model fitting, representing the percentage that can be interpreted as a regression relationship in the total change of the target variable. The closer the value is to 1, the better the fitting effect. The Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) can evaluate the complexity of the model being estimated and the quality of the model fitting data. The smaller the value, the better the model effect. Using store information, real-time weather, and eight types of geographic environmental indicators as independent variables, and customer count data as the dependent variable, a linear regression model was constructed. After adding each of the eight geographic environmental features to the evaluation model, R 2 The method for evaluating stores using indices related to the store's surrounding environment as described in Appendix 8, characterized in that the traffic index improves, the effect of the traffic index is most evident, followed by the entertainment index and the travel index, the interpretability of the target variable - changes in customer numbers is enhanced, the ACI and BIC of each index decrease, the model fitting is good, and no overfitting phenomenon occurred.
[0040] (Note 10) In S5, the SSTI index was applied to a real business scenario, and based primarily on the verification results described above, the brand was provided with a certain degree of business guidance and reference value regarding store address selection and store grouping. The method for evaluating stores using the relevant indices for the surrounding environment of a store, as described in Appendix 1, is characterized in that, if other similar conditions exist, new store addresses are given priority in terms of convenient transportation locations with some entertainment facilities, and in accordance with the demands of the real-world scenario, the store's spatiotemporal index is an objective numerical basis for commercial users to quantify the surrounding environment and rapidly improves the efficiency of geographic evaluation work.
Claims
1. This method evaluates stores using indices related to the surrounding environment, primarily by quantifying geographical environment themes that have business value and are highly interpretable, based on geographical information. The combination of geographical information and related natural features is used as an indicator score to evaluate the store's management status or capacity to meet management needs, providing a basis for reference in the store's business scenario. Step S1 involves collecting geographic information data around offline stores, Step S2 involves statistically modeling the entities surrounding the store, Step S3 involves rationally extracting the theme characteristic index SSTI, Step S4 involves constructing a linear regression model evaluation system, A method for evaluating a store using an index related to the surrounding environment of a store, characterized by including step S5, which involves applying the index SSTI to a real business scenario.
2. In S1 above, geographic information data around the offline store is collected by mainly using different data sources, thereby obtaining geographic information around the target store, mainly including information about the surrounding environment and related natural features. The method for evaluating a store using an index of the surrounding environment of a store, as described in claim 1, is characterized in that, after acquiring geographic information, facilities near the target store are divided into eight categories: transportation, residence, shopping, gourmet food, travel, lifestyle, work, and entertainment, and for each category, the number of facilities related to that category near the target store and the distance information from the target store are obtained.
3. In S2 above, when statistically modeling the entities surrounding the store, a theme intensity model and a theme coherence model are mainly established, and by combining the results of the two models, a large-scale spatiotemporal index under the theme is generated. The method for evaluating stores using indexes related to the store's surrounding environment, as described in claim 1, is characterized in that it primarily uses a linear regression model to verify the magnitude of the impact of large-scale spatiotemporal indicators on the number of customers, and if the verification is successful, the large-scale spatiotemporal indicators can be applied to the brand's real-world business scenarios, contributing to an improvement in the efficiency of the brand's evaluation of the store's surrounding environment, and ultimately serving as an important reference for demand such as store address selection.
4. In S2 above, when performing statistical modeling, the store's geographical location environment theme score is influenced by the number intensity and degree of clustering of themes of interest around the store. Using the POIs around the store as a sample, the following model is established for the distribution of themes POIs around the store: [Math 1] Here, 's' is a superscript for store. 't' is a subscript for themes, encompassing eight categories: transportation, housing, shopping, dining, travel, lifestyle, work, and entertainment. Loc(POI ｔ （ｓ） ) represents the distribution of POI themes around stores, [Math 2] This is a store-around theme POI intensity model that follows a Bernoulli distribution, p ｔ （ｓ） This is a probability parameter of a store-related theme POI intensity model that conforms to a beta distribution, α ｔ , β ｔ These are variables in the intensity parameter distribution model, [Math 3] This is a store-related theme POI set model that follows a two-dimensional normal distribution, Σ ｔ （ｓ） This is the variance of the store-related theme POI set model, which matches a Gaussian distribution, and [Math 4] satisfies, (σ ｔ （ｓ） ) ２ is the dispersion of the POI longitude or dimension, μ ｔ , λ ｔ The method for evaluating a store using an index related to the surrounding environment of a store, as described in claim 3, wherein is a variable of a collective parameter distribution model.
5. In S2, the model-based calculation process includes inputs, outputs, and calculation steps, and specifically, it is as follows: Input: t: Evaluation themes including a total of eight categories: transportation, housing, shopping, dining, travel, lifestyle, work, and entertainment. m: Total number of stores to be evaluated, n ｔ （ｓ） : Number of POIs searched in s stores under theme t, N ｔ （ｓ） : Number of POI queries for s stores under theme t, p ｔ （ｓ） Initial probability parameters of the POI strength model for s stores under theme t, σ ｔ （ｓ） : Standard deviation of all POIs and store distances for s stores under theme t, (x ｉ （ｓ） , y ｉ （ｓ） ): This is the coordinate of the iPOI point near the store relative to the store. α and β are the inter-store strength model priors, following a beta distribution, and are model variables. μ and λ are prior variables of the inter-store settling ratio model, following a Gaussian distribution, and are model variables. Output: SSTI ｔ （ｓ） A method for evaluating stores using an index related to the surrounding environment of stores, as described in feature 3, which is the score of s stores under theme t.
6. The calculation process in S2 is as follows: 1) Execute at each t, 2) Start the loop, 3) In the intensity model, the intensity of themes POI around stores follows a Bernoulli distribution, and its probability parameter follows a beta distribution. The integrated probability density function of the intensity model is as follows: [Math 5] 4) Based on maximum likelihood estimation, [Math 6] Let the partial derivative of with respect to p be 0, and then set p to p ｔ Here, the probability parameter of the intensity model for each store is p ｔ （ｓ） And, 5) In the accretion model, the store-related theme POI follows a two-dimensional normal distribution, the variance of the POI accretion model matches a Gaussian distribution, and the combined probability density function of accretion is as follows: [Number 7] 6) Based on maximum likelihood estimation, [Number 8] Let the partial derivative of with respect to σ be 0, ｔ σ ｔ Here, the standard deviation of the concentration model for each store is σ ｔ （ｓ） And, 7) Probability parameter p of the theme intensity model under each store's t-theme ｔ （ｓ） and the standard deviation σ of the concentration model ｔ （ｓ） Obtained, 8) End the loop, 9) A linear regression model was constructed using store information and real-time weather data as independent variables, and customer count data as the dependent variable. 10) R ２ <- Model determination coefficient, 11) AIC <- Akaike Information Criterion, 12) BIC <- Bayesian Information Standard, 13) Theme scoring formula SSTI ｔ （ｓ） = p ｔ （ｓ） / σ ｔ （ｓ） Define, 14) Execute at each t, 15) Start the loop, 16) Theme points for each store's theme SSTI ｔ （ｓ） Obtained, 17) Thematic score features SSTI in linear regression models ｔ （ｓ） Add, 18) R after theme points are added ２ ｔ ACI ｔ and BIC ｔ Obtained, 19) End the loop, 20) R below each t ２ ｔ The effect does not clearly improve or ACI ｔ It is too big or BIC ｔ If it is too large, 21) Execute the loop, 22) Redefine the formula for calculating theme scores and perform steps 14-19. 23) R ２ ｔ ACI ｔ BIC ｔ Observe, 24) R at each t ２ ｔ The effect of ACI has clearly improved and ｔ BIC ｔ Until all the numbers are reasonable, 25) End the loop, 26) The score SSTI of the corresponding t ｔ （ｓ） Hold, 27) SSTI ｔ （ｓ） The method for evaluating stores using an index related to the surrounding environment of a store, as described in claim 5, is characterized in that it can be applied to real business scenarios such as selecting addresses and grouping brand stores, and provides the basis for scoring themes in each dimension so that evaluators can select the optimal geographical environment.
7. In S3, the rational extraction of the theme characteristic index SSTI is mainly done using maximum likelihood estimation for each store and each theme. [Number 9] and [Number 10] This involves evaluating, in particular, [Math 11] The larger the value, the greater the numerical strength of the thematic POIs around the store. [Math 12] As the size increases and the concentration of themed POIs around the store increases, the following SSTI ｔ （ｓ） Define and describe the POI distribution under the store theme, SSSI ｔ （ｓ） =p ｔ （ｓ） / σ ｔ （ｓ） Here, [Number 13] This is the probability parameter of the POI intensity model under the t theme around s stores estimated by MLE, [Number 14] This is the standard deviation of the POI set model under the t theme around s stores, estimated by MLE, and represents the SSTI of the stores. ｔ （ｓ） A method for evaluating stores using an index related to the surrounding environment of a store, as described in item 1, characterized in that the higher the number, the clearer the thematic characteristics become.
8. In S4, the linear regression model evaluation system specifically selects indicators that can be used for monitoring and that can evaluate the store management status or the target of management, then establishes a model using geographic information indicators and other relevant natural features, and embodies the rationality of the geographic information indicators through the model fitting degree gain. The linear regression model has good linear interpretability and feature compatibility, possesses a complete quantified feature evaluation and inference mechanism, can easily describe the gain of feature superposition, has a natural fit for evaluating large-scale spatiotemporal features of offline stores, and does not require preconditions for training and test sets, so the linear regression model evaluates the magnitude and positive / negative influence of independent variable features on the dependent variable, as described in claim 1, for store evaluation methods using indexes related to the store's surrounding environment.
9. In S4, the coefficient of determination is used as an index to evaluate the degree of model fitting, representing the percentage that can be interpreted as a regression relationship in the total change of the target variable. The closer the value is to 1, the better the fitting effect. The Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) can evaluate the complexity of the model being estimated and the quality of the model fitting data. The smaller the value, the better the model effect. Using store information, real-time weather, and eight types of geographic environmental indicators as independent variables, and customer count data as the dependent variable, a linear regression model was constructed. After adding one of the eight types of geographic environmental features to the evaluation model, R ２ The method for evaluating a store using indices related to the store's surrounding environment, as described in 8, is characterized in that the effect of the traffic index is most evident, followed by the entertainment index and the travel index, the interpretability of the target variable - changes in the number of customers is enhanced, the ACI and BIC of each index all decrease, the model fitting is good, and no overfitting phenomenon occurred.
10. In S5, the SSTI index is applied to real business scenarios, and based primarily on the verification results described above, the brand is provided with a certain degree of business guidance and reference value regarding store address selection and store grouping. The method for evaluating a store using an index related to the store's surrounding environment, as described in claim 1, is characterized in that, if other similar conditions exist, the new store address is given priority in terms of convenient transportation locations with some entertainment facilities, and in accordance with the demands of the real-world scenario, the store's large-scale spatiotemporal index is an objective numerical basis for commercial users to quantify the surrounding environment and rapidly improves the efficiency of geographic evaluation work.