Shop site selection recommendation method and system based on big data

Abstract

The invention relates to the technical field of data processing, in particular to a shop site selection recommendation method and system based on big data. The method comprises the following steps: acquiring positions of a plurality of candidate shops and passenger flow data of a plurality of points of interest; for any candidate shop position, determining the passenger flow toughness strength between the point of interest and the candidate shop; performing clustering processing on the passenger flow toughness strength to obtain a plurality of commercial clusters; performing link breaking processing on association between points of interest in the business cluster and candidate shops, and ensuring that each point of interest is associated with at most one candidate shop; calculating the collaborative stability of the business cluster after the link breaking processing, and constructing an objective function of an ant colony optimization algorithm based on the collaborative stability; and performing optimization processing on the candidate shops by using an ant colony optimization algorithm to obtain a final shop site selection result. The method has the effect of improving the accuracy and reliability of the shop site selection result.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method and system for recommending store locations based on big data. Background Technology

[0002] Store location selection is one of the most important strategic decisions in industries such as retail and catering. A scientifically sound location selection plan can not only help businesses acquire sufficient and stable customer traffic, but also reduce market entry risks and improve long-term profitability. Conversely, an inappropriate location may lead to insufficient customer traffic, decreased profitability, or even the risk of business failure.

[0003] In the traditional model, businesses typically rely on managers' experience, intuition, or limited market research data for site selection. While this approach has some practical applicability, it has significant limitations: First, relying on experience makes it difficult to systematically cover multi-dimensional factors, such as geographic distribution, transportation convenience, consumer characteristics, and surrounding competitive relationships. Second, limited research data struggles to reflect the dynamic and temporal characteristics of customer flow, leading to biased and uncertain evaluation results. Therefore, traditional site selection methods often exhibit strong subjectivity and randomness, making it difficult to ensure long-term scientific rigor and stability in a highly competitive market environment. With the development of big data technology, utilizing massive amounts of multi-dimensional data for quantitative analysis and intelligent decision-making of candidate store locations has become an important means to improve the scientific rigor and accuracy of site selection. Store location selection is essentially a complex optimization problem integrating geographical, commercial, and social multi-dimensional heterogeneous data, and its candidate solution space typically exhibits high dimensionality and multi-peak characteristics. For such path optimization problems, existing technologies often employ ant colony optimization algorithms for solving them. This algorithm, by simulating ant colony foraging behavior, possesses strong global search capabilities.

[0004] However, directly applying the original ant colony optimization algorithm in store location selection scenarios has obvious drawbacks. Its objective function is usually simply set to maximize customer traffic or minimize cost. This setting ignores the synergistic effect of commercial clusters formed between different points of interest, and also fails to fully consider the potential resource competition between multiple candidate stores for the same customer source, thus affecting the accuracy and reliability of the final location recommendation results. Summary of the Invention

[0005] To improve the accuracy and reliability of store location selection results, this application provides a store location recommendation method and system based on big data.

[0006] Firstly, this application provides a store location recommendation method based on big data, employing the following technical solution: A big data-based store location recommendation method includes: acquiring the locations of multiple candidate stores and customer traffic data of multiple points of interest (POIs); for any candidate store location, determining the customer flow resilience strength between the POI and the candidate store based on the customer traffic data of the POIs and the route distance from the POI to the candidate store; clustering the customer flow resilience strength to obtain multiple business clusters; breaking the links between the POIs and candidate stores within the business clusters to ensure that each POI is associated with at most one candidate store; calculating the collaborative stability of the business clusters after the link breaking process, and constructing the objective function of an ant colony optimization algorithm based on the collaborative stability; and using the ant colony optimization algorithm to optimize the candidate stores to obtain the final store location result.

[0007] First, a customer flow resilience index was established to comprehensively quantify the actual attractiveness of candidate stores to points of interest. This index considers three factors simultaneously: customer flow scale, time stability, and route distance, thus avoiding the simplistic evaluation of traditional methods that rely solely on total customer flow or distance. This makes the measurement of candidate sites more in line with commercial realities. Based on customer flow resilience, the relationships between all candidate stores and points of interest are clustered, identifying synergistic effects between multiple points of interest and candidate stores, forming commercial clusters. This allows the model to not only capture the value of individual points of interest but also reflect the brand-attracting effect of regional clusters on stores, thereby ensuring stronger globality and stability in site selection decisions. Furthermore, this method introduces a chain-breaking mechanism to constrain the association between points of interest and candidate stores within commercial clusters, ensuring that each point of interest can correspond to at most one candidate store. This measure effectively avoids internal competition among multiple candidate stores surrounding the same point of interest. Furthermore, this application evaluates the internal quality of each cluster by calculating the collaborative stability of the commercial clusters after link breakage processing. High collaborative stability means that the points of interest and candidate stores within the commercial cluster have a strong and relatively consistent commercial radiation capability, thus bringing more stable and predictable customer flow revenue to enterprises and reducing the operational risk of site selection. Finally, collaborative stability is introduced as the objective function into the ant colony optimization algorithm, enabling the optimization process to prioritize the retention of high-quality store-point of interest combinations in the global search, thereby efficiently converging to the store location scheme with the highest overall commercial value in the large-scale solution space. This effectively improves the accuracy and reliability of store location results.

[0008] Optionally, for any point of interest, the visitor flow data can be obtained by calling the crowd heat map or visitor flow statistics interface provided by the map API to obtain the visitor flow data within a preset range around the point of interest.

[0009] By calling the crowd heatmap or passenger flow statistics interface provided by the map API, passenger flow data is obtained, achieving automation, scalability, and dynamism in data acquisition. This reduces the cost and time of data collection.

[0010] Optionally, the step of determining the resilience of customer flow between the point of interest and the candidate store includes: dividing the customer flow data of any point of interest into multiple time periods, each time period including multiple customer flow data collected; arranging the customer flow data collected at the same time in different time periods in sequence to form a corresponding customer flow sequence; using the ratio of the mean to the standard deviation of the customer flow sequence as a local stability index; using the sum of the local stability indices corresponding to each customer flow sequence as a customer flow stability index; and using the product of the customer flow stability index and the reciprocal of the route distance from the point of interest to the candidate store as the customer flow resilience.

[0011] By combining the stability of customer traffic patterns (the ratio of mean to standard deviation) with the physical distance to the store (the reciprocal of the route distance), a deeper assessment of the quality of a customer base can be achieved. This effectively distinguishes customer sources with similar traffic volumes but different stability, such as the stable traffic flow in office buildings versus the pulsating traffic flow in concert venues.

[0012] Optionally, the K-means clustering algorithm can be used to cluster the passenger flow resilience intensity.

[0013] Optionally, the elbow method can be used to determine the number of clusters in the K-means clustering algorithm.

[0014] The elbow method strikes a balance between too few and too many clusters, ensuring that the resulting business clusters maintain high internal cohesion while also exhibiting good separability between different clusters. Optionally, the method for determining the collaborative stability of the business cluster includes: for any passenger flow resilience strength, calculating the absolute value of the difference between the passenger flow resilience strength and the mean collaborative stability in the business cluster, taking the ratio of the passenger flow resilience strength to the absolute value as the local stability, and taking the mean of the corresponding multiple local stability in the business cluster as the collaborative stability.

[0015] By calculating the dispersion of passenger flow resilience strength and its mean within a business cluster, an index for measuring the quality and consistency within the business cluster, namely, collaborative stability, is constructed.

[0016] Optionally, the objective function can be constructed by using the sum of the collaborative stability of all business clusters after the chain break as the objective function.

[0017] By using the sum of the collaborative stability of all business clusters as the final objective function of the ant colony algorithm, it is ensured that the algorithm always moves towards building an overall store network composed of multiple high-quality, highly collaborative business clusters when performing a global search, thus obtaining a more robust and efficient site selection result in terms of macro-layout.

[0018] Optionally, in the ant colony optimization algorithm, the disconnection between all points of interest and candidate stores is represented as a one-dimensional vector, and the one-dimensional vector is used as the position of the ant.

[0019] Optionally, the final store location result includes: disconnecting the links between the points of interest and candidate stores corresponding to the elements with a value of 0 in the binary vector output by the ant colony optimization algorithm, and taking the candidate stores that are still associated with the points of interest after the disconnection as the final store location result.

[0020] Secondly, this application provides a store location recommendation system based on big data, employing the following technical solution: The big data-based store location recommendation system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the aforementioned big data-based store location recommendation method.

[0021] The above-mentioned big data-based store location recommendation method is generated into a computer program and stored in memory so that it can be loaded and executed by the processor. Thus, a system is created based on the memory and processor for easy use.

[0022] This application has the following technical advantages: This method constructs a customer flow resilience index to comprehensively evaluate customer quality from three dimensions: scale, stability, and distance of points of interest. It also establishes a business clustering and chain breakage handling mechanism to address internal resource competition in multi-store layouts. Furthermore, it constructs an optimization objective function centered on collaborative stability to guide the algorithm in finding a store network with optimal overall synergy, thereby making the site selection results more accurate and reliable. Attached Figure Description

[0023] Figure 1 This is a flowchart of the store location recommendation method based on big data in the embodiments of this application. Detailed Implementation

[0024] This application discloses a big data-based store location recommendation method. First, it acquires information on multiple pre-selected candidate stores and points of interest (POIs), analyzes the impact of POIs on candidate stores, and constructs a customer flow resilience strength. Based on the customer flow resilience strength, it performs clustering to obtain multiple business clusters. It then performs disconnection processing between POIs and candidate stores, ensuring that each POI is associated with only one candidate store. Based on the disconnected business clusters, it calculates the collaborative stability of the corresponding business clusters and constructs an ant colony algorithm objective function to optimize the candidate stores, obtaining the final location selection.

[0025] Reference Figure 1 The big data-based store location recommendation method includes steps S1-S5.

[0026] S1: Obtain the locations of multiple candidate stores and the customer traffic data of multiple points of interest.

[0027] Based on the company's business intentions and strategic plans, several candidate store locations are initially selected within the target area. These locations may include shops available for rent or sale, or reserved spaces within commercial complexes.

[0028] Then, using map API (Application Programming Interface) services, such as Baidu Maps API or Gaode Maps API, the precise latitude and longitude coordinates of these N candidate store locations are collected programmatically. Simultaneously, to assess the customer traffic potential of the candidate stores, it is necessary to identify and collect all potential customer traffic sources within target area A, i.e., Points of Interest (POIs). POIs include geographical entities that typically attract large crowds, such as residential areas, schools, large shopping malls, supermarkets, stadiums, transportation hubs, and office buildings. The latitude and longitude coordinates of these POIs are also obtained through the map API interface.

[0029] Historical visitor traffic data is collected around each point of interest (POI). For any POI, the visitor traffic data is divided into multiple time periods, each containing multiple collected visitor traffic data points. In this embodiment, by calling the crowd heatmap or visitor traffic statistics interface provided by the map API, visitor traffic data within a 1-kilometer radius of all POIs in region A over the past month is collected. To capture the temporal variation patterns of visitor traffic, the data collection interval can be set to 1 hour. Alternatively, the time period can be set to 1 day, with 24 visitor traffic data points collected daily within that period; in this embodiment, visitor traffic data is collected continuously for 30 days. It is understood that the scope and time interval of visitor traffic data collection can be adjusted according to specific business types and required analytical precision.

[0030] S2: For any candidate store location, determine the passenger flow resilience between the point of interest and the candidate store based on the passenger flow data of the point of interest and the route distance from the point of interest to the candidate store.

[0031] A high-value point of interest not only provides a large volume of foot traffic, but more importantly, it provides a stable flow of customers. For example, a large office building offers a highly consistent flow of white-collar workers on weekdays, while a popular concert venue provides a burst of traffic that is highly unpredictable and sporadic. For most business models, stable foot traffic translates to predictable sales and helps businesses make precise arrangements for inventory management and staffing. Therefore, the commercial value of locating a shop near an office building is usually far greater than that near a concert venue.

[0032] Based on this, this step aims to construct a comprehensive indicator that can simultaneously quantify distance, customer flow scale, and customer flow stability, namely, customer flow resilience, to reflect the actual attractiveness of a candidate store to a point of interest.

[0033] The steps for determining the resilience of customer flow between the point of interest and the candidate store include: arranging customer flow data collected at the same time in different time periods sequentially to form a corresponding customer flow sequence; using the ratio of the mean to the standard deviation of the customer flow sequence as a local stability index; using the sum of the local stability indices corresponding to each customer flow sequence as a customer flow stability index; and using the product of the customer flow stability index and the reciprocal of the route distance from the point of interest to the candidate store as the customer flow resilience.

[0034] The formula for calculating passenger flow resilience can be expressed as: In the formula, Indicates the first The point of interest and the first The resilience of customer traffic for each candidate store. Indicates the first The point of interest and the first The straight-line distance between the candidate stores This indicates the number of times passenger flow data is collected within a preset time period. In this embodiment, the time period is 1 day, and data is collected 24 times per day. Passenger flow data from the same collection time within multiple time periods constitute the total passenger flow data. Indicates the first The first point of interest The mean of the passenger flow sequence formed by the passenger flow data collected in each iteration; Indicates the first The first point of interest The standard deviation of the passenger flow series composed of the passenger flow data collected in each iteration This represents a hyperparameter used to avoid the denominator being zero, which would prevent calculation. In this application, the value is 0.1, but in other embodiments, it can be selected based on the experience of those skilled in the art.

[0035] When the The point of interest and the first The shorter the route distance between the candidate stores, the greater the resilience of customer flow, indicating that it is more convenient for customers to reach the stores from their points of interest. Meanwhile, when the... The larger the mean and the smaller the standard deviation of the visitor flow sequence for each point of interest, the better. The larger the value, the more stable the passenger flow; this indicates that the point of interest can provide a large-scale and stable passenger flow. In summary, the greater the passenger flow resilience, the more stable the point of interest. The point of interest for the first The stronger the commercial reach of a candidate store, the higher the quality of its customer traffic.

[0036] S3: Cluster the passenger flow resilience to obtain multiple business clusters; break the association between points of interest and candidate stores within the business clusters to ensure that each point of interest is associated with at most one candidate store.

[0037] A feature matrix is ​​constructed by comparing the customer flow resilience between all points of interest and all candidate stores. This matrix is ​​then used as input data for the K-means clustering algorithm, and the Elbow Method is employed to determine the optimal number of clusters. The elbow method calculates different... Calculate the sum of squared errors (SSE) within the cluster at the specified value, and plot the K-SSE curve, selecting the elbow corresponding to the value on the curve. The value represents the optimal number of clusters, signifying the balance point where the diminishing returns from increasing the number of clusters are most significant. Both the K-means clustering algorithm and the elbow method are well-known techniques in the field and will not be elaborated upon here. After clustering, each cluster is defined as a business cluster, containing several pairs of points of interest—candidate shops—that exhibit similar resilience in customer traffic.

[0038] When selecting locations for multiple stores, a key issue is avoiding competition between different branches for the same customer traffic, i.e., internal competition. To address this problem, this application introduces a chain-breaking mechanism and constructs a new objective function to guide the search direction of the ant colony optimization algorithm.

[0039] In S3, multiple business clusters were obtained through clustering. However, the same point of interest may exhibit high customer flow resilience with multiple different candidate stores, and thus be assigned to the same or different business clusters. This physically means that multiple candidate stores are competing for the same customer flow source.

[0040] When evaluating a set of site selection options, a constraint is enforced that a point of interest can serve at most one candidate store. Specifically, this is simulated by randomly disconnecting links between a point of interest and a candidate store. For example, disconnecting a point of interest... with candidate stores The link does not contain any point of interest. with candidate stores The resilience of passenger flow will correspond to the points of interest in the commercial cluster. with candidate stores The customer flow resilience strength between points of interest and candidate stores is removed. At this point, the number of customer flow resilience strength values ​​for all points of interest and candidate stores is... .

[0041] S4: Calculate the collaborative stability of the commercial cluster after the link breakage process, and construct the objective function of the ant colony optimization algorithm based on the collaborative stability.

[0042] To measure the internal quality of each business cluster after the chain break, this application constructs a business cluster collaborative stability index. A high-quality business cluster should have generally high customer flow resilience strength among each point of interest-candidate store pair, and the strength values ​​should be relatively concentrated with small differences.

[0043] The method for determining the collaborative stability of the commercial cluster includes: for any passenger flow resilience strength, calculating the absolute value of the difference between the passenger flow resilience strength and the mean collaborative stability in the commercial cluster, taking the ratio of the passenger flow resilience strength to the absolute value as the local stability, and taking the mean of the corresponding multiple local stability in the commercial cluster as the collaborative stability.

[0044] For any business cluster, the formula for calculating its collaborative stability can be expressed as: In the formula, Indicates the first Stability of business cluster collaboration Indicates the first The number of customer flow resilience values ​​in a business cluster Indicates the first The first business cluster The value of passenger flow resilience. Indicates the first The mean of passenger flow resilience in each cluster. This represents a hyperparameter used to avoid situations where the denominator is zero, which would prevent calculation. In this application, the value is 0.1, but it can be selected according to the situation.

[0045] If a commercial cluster belongs to a high-quality cluster area, then the points of interest within that cluster should have a strong commercial radiation capacity between them and the candidate shops, corresponding to a high degree of customer flow resilience. At the same time, the commercial radiation capacity of different points of interest to the candidate shops should be relatively similar. The smaller the value of the cluster, the higher its collaborative stability and the stronger its internal cohesion. This results in lower location risk when selecting shops within the cluster, leading to a higher calculated collaborative stability for the commercial cluster. Conversely, if a commercial cluster is of low quality, the points of interest within it have a weaker commercial reach to candidate shops. The items are relatively small, and the commercial reach of candidate stores varies greatly depending on the different points of interest. The larger value results in a lower calculated stability of the business cluster collaboration.

[0046] The sum of the collaborative stability of all business clusters after the chain breakage is used as the objective function.

[0047] S5: Use the ant colony optimization algorithm to optimize the candidate stores and obtain the final store location results.

[0048] In ant colony optimization algorithms, the position of each ant represents a potential solution. In this application, a solution is a combination of all possible connections between points of interest and candidate shops. Specifically, all possible connections (total...) can be considered. One, of which For the number of points of interest, (The number of candidate stores) is represented as a string of length . This is a one-dimensional binary vector. Each element in the vector is either 1 or 0, representing whether the corresponding point of interest-candidate store connection is retained or broken. This vector represents the ant's position.

[0049] The optimization process of the algorithm is as follows: Initialize a group of ants and randomly place them in the solution space (i.e., randomly generate initial binary vectors). Each ant calculates the objective function based on its current position (i.e., the connection combination it represents). The ant colony algorithm determines the next ant's movement direction (i.e., modifying certain values ​​in its binary vector) based on the objective function value and pheromone concentration. Paths with higher objective function values ​​attract more ants as their pheromone concentration increases faster. This process is iterated until the algorithm converges or reaches the preset maximum number of iterations. The optimization process of the ant colony algorithm is a well-known technique and will not be elaborated here. After convergence, the algorithm outputs the optimal "ant position," which is the position that maximizes the objective function value. The length that reaches the maximum value is A one-dimensional binary vector. The "point of interest - candidate store" connections corresponding to elements with a value of 0 in this vector are broken. Finally, in all... Of the candidate stores, only those that still have a connection with at least one point of interest in the optimal solution are retained. These retained stores constitute the final recommended store location results.

[0050] This application also discloses a big data-based store location recommendation system, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the big data-based store location recommendation method according to this application.

[0051] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0052] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A store location recommendation method based on big data, characterized in that, include: Acquire the locations of multiple candidate stores and the customer traffic data of multiple points of interest; for any candidate store location, determine the customer traffic resilience between the point of interest and the candidate store based on the customer traffic data of the point of interest and the route distance from the point of interest to the candidate store; Clustering the passenger flow resilience intensity yields multiple business clusters; The association between points of interest and candidate stores within the business cluster is broken to ensure that each point of interest is associated with at most one candidate store. Calculate the collaborative stability of the commercial cluster after the link breakage treatment, and construct the objective function of the ant colony optimization algorithm based on the collaborative stability. The candidate stores are optimized using the ant colony optimization algorithm to obtain the final store location results.

2. The store location recommendation method based on big data according to claim 1, characterized in that, For any point of interest, the visitor flow data can be obtained by calling the crowd heat map or visitor flow statistics interface provided by the map API, and then the visitor flow data within a preset range around the point of interest can be obtained.

3. The store location recommendation method based on big data according to claim 1, characterized in that, The steps for determining the resilience of customer flow between the point of interest and the candidate store include: dividing the customer flow data of any point of interest into multiple time periods, with each time period including multiple customer flow data collected; arranging the customer flow data collected at the same time in different time periods sequentially to form a corresponding customer flow sequence; using the ratio of the mean to the standard deviation of the customer flow sequence as a local stability index; using the sum of the local stability indices corresponding to each customer flow sequence as a customer flow stability index; and using the product of the customer flow stability index and the reciprocal of the route distance from the point of interest to the candidate store as the customer flow resilience.

4. The store location recommendation method based on big data according to claim 1, characterized in that, The K-means clustering algorithm was used to cluster the passenger flow resilience intensity.

5. The store location recommendation method based on big data according to claim 4, characterized in that, The elbow method was used to determine the number of clusters in the K-means clustering algorithm.

6. The store location recommendation method based on big data according to claim 1, characterized in that, The method for determining the collaborative stability of the business cluster includes: for any passenger flow resilience strength, calculating the absolute value of the difference between the passenger flow resilience strength and the mean collaborative stability in the business cluster, taking the ratio of the passenger flow resilience strength to the absolute value as the local stability, and taking the mean of the corresponding multiple local stability in the business cluster as the collaborative stability.

7. The store location recommendation method based on big data according to claim 1, characterized in that, The method for constructing the objective function includes: using the sum of the collaborative stability of all business clusters after the chain breakage process as the objective function.

8. The store location recommendation method based on big data according to claim 1, characterized in that, In the ant colony optimization algorithm, the disconnection between all points of interest and candidate stores is represented as a one-dimensional vector, and the one-dimensional vector is used as the position of the ant.

9. The store location recommendation method based on big data according to claim 1, characterized in that, The final store location selection result includes: disconnecting the links between the points of interest and candidate stores corresponding to the elements with a value of 0 in the binary vector output by the ant colony optimization algorithm, and taking the candidate stores that are still associated with the points of interest after disconnection as the final store location selection result.

10. A store location recommendation system based on big data, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement the big data-based store location recommendation method according to any one of claims 1-9.

Images