A Location Selection Method for Distribution Centers Considering Potential Pharmaceutical Retailers

Abstract

The present invention relates to a method for selecting the location of a distribution center considering potential pharmaceutical retailers, belonging to the technical field of location selection, and comprising the following steps: S1: Obtain the details of existing customer orders of a pharmaceutical logistics enterprise and mine potential customer data; S2: Construct an evaluation index system for the conversion probability of potential pharmaceutical retailers; S3: Use the entropy weight method to calculate the comprehensive evaluation result and perform normalization processing on it to obtain the conversion probability of potential customers; S4: Define the coverage range of the distribution center and screen out the customer points that can be delivered by the central warehouse; S5: Use the DBSCAN algorithm to remove remote points; S6: Perform weighted K-means clustering with the conversion probability of potential customers as the weight; S7: Use the centroid method to find the location of the distribution center in each clustering area respectively.

Description

Technical Field

[0001] The present invention belongs to the technical field of site selection, and relates to a method for selecting a distribution center considering potential pharmaceutical retailers. Background Art

[0002] Facing a large-scale customer group, as the customer group grows, the original warehouse of pharmaceutical logistics enterprises is difficult to meet the distribution requirements. To save logistics costs and improve distribution efficiency, it is necessary to establish a distribution center to distribute to pharmaceutical retailers. When studying the site selection problem of the distribution center, not only existing customers but also potential pharmaceutical retailers of the enterprise need to be considered for comprehensive analysis. Potential pharmaceutical retailers refer to customers who have a demand for services such as drug wholesaling and distribution provided by a certain pharmaceutical logistics enterprise, and they have a certain purchasing power but have not yet become real customers of the enterprise. It is a difficult point for the enterprise to estimate the probability of each potential customer converting into a real customer of the enterprise based on the available data. In addition, after considering potential customers, how to select the site of the distribution center for tens of thousands of scattered customers is also a practical problem.

[0003] Most existing studies are focused on the site selection of distribution centers for existing customers, rarely considering the role played by potential pharmaceutical retailers in the site selection of distribution centers and ignoring this important indicator. When studying the site selection for a large-scale customer base of tens of thousands, many scholars have proposed a solution idea of clustering first and then site selection. Lin Li et al. calculated the feature weights using the Softmax and Sigmoid regression functions according to each feature difference degree, assigned different contribution degrees to the features, and proposed the LWK-means algorithm; Ma Zongbiao et al. proposed feature-weighted fuzzy clustering based on the FCM algorithm. However, most scholars only use the K-means clustering algorithm to partition and cluster the customer locations. Even when potential customers are considered, they are only regarded as existing customers, without considering the influence of the weights of different customers on K-means clustering. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method for selecting a distribution center considering potential pharmaceutical retailers. First, it is necessary to mine the potential customer data of pharmaceutical logistics enterprises, then construct an evaluation index system for the conversion probability of potential pharmaceutical retailers from three aspects: customer value of potential pharmaceutical retailers, comprehensive strength of pharmaceutical logistics enterprises, and market role. Then, use the entropy weight method to calculate the weights of each index and the final comprehensive score, and perform normalization processing on the score to obtain the probability of each potential pharmaceutical retailer converting into a real customer. Then, after screening out the customer points that can be delivered by the central warehouse and remote customer points, use the weighted K-means clustering algorithm to cluster existing customers and potential customers. Finally, use the centroid method to select the site of the distribution center for each clustering area.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A distribution center location selection method considering potential pharmaceutical retailers, comprising the following steps:

[0007] S1: Obtain the details of existing customer orders of a pharmaceutical logistics enterprise, and mine potential customer data;

[0008] S2: Construct an evaluation index system for the conversion probability of potential pharmaceutical retailers;

[0009] S3: Use the entropy weight method to calculate the comprehensive evaluation result, and perform normalization processing on it to obtain the conversion probability of potential customers;

[0010] S4: Define the coverage area of the distribution center, and screen out the customer points that can be delivered by the central warehouse;

[0011] S5: Use the DBSCAN algorithm to remove remote points;

[0012] S6: Perform weighted K-means clustering with the conversion probability of potential customers as the weight;

[0013] S7: Use the centroid method to find the location of the distribution center in each clustering area respectively.

[0014] Furthermore, in the step S1, the existing customer information in the database of a pharmaceutical logistics enterprise includes: customer name customer, customer number customerID, customer address customerArea, customer longitude and latitude customerPosition. The process of mining potential pharmaceutical retailer data of the enterprise in the business radiation area by using the descendant collector is as follows:

[0015] S11: Select the target city served by the logistics enterprise in the map software;

[0016] S12: Input the key store names served by the logistics enterprise, including pharmacies, drugstores, clinics, health centers, and hospitals;

[0017] S13: Collect the store name, address and phone number information;

[0018] S14: Save and export the collected data;

[0019] S15: Compare the mined data with the existing customer data of the enterprise, screen out the list of potential customers to be developed, and convert all customer addresses into longitude and latitude.

[0020] Further, in the step S2, the customer value of potential pharmaceutical retailers, the comprehensive strength of pharmaceutical logistics enterprises, and the market role are used as secondary indicators. Then, the corresponding tertiary indicators are sorted out according to the secondary indicators, an evaluation index system for the conversion probability of potential pharmaceutical retailers is built, and the data to be collected is determined.

[0021] Further, in the step S3, it specifically includes the following steps:

[0022] S31: Normalize the values of the tertiary indicators. First, distinguish them into positive indicators and negative indicators according to the nature of the indicators;

[0023] The calculation formula for positive indicators is:

[0024]

[0025] The calculation formula for negative indicators is:

[0026]

[0027] Where x ij represents the original value of the i-th customer under the j-th indicator, and n represents the number of all potential customers;

[0028] S32: Calculate the proportion P ij :

[0029]

[0030] Where m represents the number of indicators, and Y ij represents the dimensionless standardized value of the i-th customer under the j-th indicator; P ij represents the proportion of the j-th indicator of the i-th customer in the sample value of this indicator, and is used as the probability when calculating the information entropy;

[0031] S33: Calculate the information entropy matrix of each potential pharmaceutical retailer customer. The calculation formula for the indicator entropy value is:

[0032]

[0033] S34: Affect the weights of the indicators of the conversion probability of potential pharmaceutical retailers by calculating the information redundancy:

[0034] d j = 1 - E j , (j = 1, 2,..., m)

[0035]

[0036] Where 1 - E j is the information entropy redundancy;

[0037] S35: Calculate the comprehensive score of each customer sample:

[0038]

[0039] S36: Normalize the comprehensive score S i Perform normalization processing.

[0040] Furthermore, in step S4, the coverage range of the distribution center is clarified, and the customer points that can be delivered by the central warehouse are screened out. The distribution ranges of the central warehouse and the distribution center should be divided according to the actual distribution requirements of the enterprise.

[0041] Furthermore, in step S5, substitute the preprocessed data into the DBSCAN clustering algorithm, adjust the two parameters of Eps and MinPts, and output the areas where distribution centers need to be established and the remote points.

[0042] Furthermore, in step S6, perform weighted K-means clustering with the potential customer conversion probability as the weight. The weight of existing customers is 1, and potential customers are calculated by substituting the values after normalization in step S36. The calculation formula is:

[0043]

[0044] Use the weighted K-means algorithm for secondary clustering, match the longitude and latitude and weights for each pharmaceutical retailer customer, and set the maximum weight for each cluster after clustering to constrain the size of the cluster area.

[0045] Furthermore, in step S7, use the centroid method to find the location of the distribution center for each clustering area respectively. First, determine the longitude and latitude of each retailer, mark them in the rectangular coordinate system, determine the Euclidean distance between each retailer, and find the centroid with the shortest sum of distances to each point within the area, which is the location point of the distribution center.

[0046] Furthermore, the centroid method has the following assumptions in the location calculation:

[0047] (1) The transportation cost is only related to the straight-line distance between the distribution center and the customer point;

[0048] (2) Do not consider the environmental factors and facade prices of the location where the distribution center is located;

[0049] Suppose there are n pharmaceutical retailer customer points, distributed at different coordinate points (x j , y j ). Assume that the distribution center is set at (x0, y0), and the total transportation cost is:

[0050]

[0051] where aj is the freight per unit weight and per unit distance from the distribution center to customer point j; w j is the transportation volume to customer point j; d j is the straight-line distance from the distribution center to customer point j:

[0052]

[0053] When selecting a distribution center, ensure that the total transportation cost is minimized.

[0054] The beneficial effects of the present invention are as follows:

[0055] 1. For large-scale retailer customers, the end logistics enterprise can use a two-stage method of clustering first and then site selection to select the location of the distribution center;

[0056] 2. When selecting a site, considering the needs of potential customers, a potential pharmaceutical retailer conversion probability evaluation index system is constructed to evaluate the value of potential customers;

[0057] 3. Design the K-means algorithm to achieve clustering under the maximum weight constraint and obtain the clustering results of the embodiments.

[0058] Other advantages, objectives and features of the present invention will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be described in detail preferably with reference to the accompanying drawings, where:

[0060] Figure 1 is a schematic diagram of the distribution center site selection method considering potential pharmaceutical retailers according to the present invention;

[0061] Figure 2 is the main algorithm flow chart of the distribution center site selection method considering potential pharmaceutical retailers according to the present invention;

[0062] Figure 3 is the clustering result diagram of the distribution center site selection method considering potential pharmaceutical retailers according to the present invention;

[0063] Figure 4 is the site selection result diagram of the distribution center site selection method considering potential pharmaceutical retailers according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0065] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0066] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0067] The present invention provides a method for selecting a distribution center site considering potential pharmaceutical retailers. The method adopts a method of clustering first and then selecting a site. The needs of potential customers are considered when selecting a site, and a reference is provided for terminal logistics companies to select a distribution center site.

[0068] Example 1: Combination Figure 1-2 The present embodiment is described as a method for selecting a distribution center location considering potential pharmaceutical retailers, including but not limited to the following steps:

[0069] S1: Obtain the order details of existing customers of a pharmaceutical logistics company and mine potential customer data;

[0070] S2: Construct a potential pharmaceutical retailer conversion probability evaluation index system;

[0071] S3: Calculate the comprehensive evaluation results using the entropy weight method, and normalize them to obtain the conversion probability of potential customers;

[0072] S4: Clarify the distribution center coverage and filter out customer points that can be delivered by the central warehouse;

[0073] S5: Use DBSCAN algorithm to remove remote points;

[0074] S6: Perform weighted K-means clustering with the conversion probability of potential customers as weight;

[0075] S7: Use the centroid method to find the distribution center position of each cluster area.

[0076] Below is an example of a pharmaceutical logistics company. The company's experience covers the retail and wholesale of various Chinese and Western medicines and medical devices. Its main customers include drug stores, pharmacies, clinics, health centers, and hospitals. Its business covers many counties and districts in the southwest region, but the company currently has only one central warehouse. The company provided the names and addresses of its customers as of 2021, the names and addresses of potential pharmaceutical retailers in the next three years, and the company's market share in various regions.

[0077] Step 1: Obtain the order details of existing customers of a pharmaceutical logistics company and mine potential customer data.

[0078] S11: Open the task website of Amap;

[0079] S12: Select the target cities that logistics companies serve;

[0080] S13: Input the key store names served by the logistics enterprise (e.g., for pharmaceutical logistics enterprises, input keywords such as drugstore, pharmacy, clinic, health room, hospital, etc.);

[0081] S14: Collect information such as store name, address and telephone number;

[0082] S15: Save and export the collected data.

[0083] Compare the mined data with the company's existing customer data to screen out a list of potential customers to be developed. At this stage, all customer addresses need to be converted into longitude and latitude to provide data support for subsequent algorithms.

[0084] The steps to convert an address to latitude and longitude are as follows:

[0085] (1) Save the geographic location data of potential customers in an Excel table, clean the data, and delete duplicate data rows;

[0086] (2) Apply for a key in the Amap console;

[0087] (3) Call the Amap API to write Python code to convert the potential customer’s geographic location into longitude and latitude and determine the horizontal and vertical coordinates of the demand point.

[0088] Step 2: Construct an evaluation index system for the conversion probability of potential pharmaceutical retailers, as shown in Table 1.

[0089] Table 1

[0090]

[0091]

[0092] The above qualitative indicators cannot be measured by accurate data and can only be used to evaluate enterprises, which are determined by expert scoring. Quantitative indicators can often be obtained through comprehensive analysis of a company's transaction data over a period of time or market surveys. The purpose of constructing this evaluation index system is as follows: The conversion probability of potential pharmaceutical retailers into real customers can be calculated based on the comprehensive scores of different potential customers. Regarding the weight of real customers as 1 and the size of the conversion probability of potential customers as the weight of potential customers, and then combining it with customer longitude and latitude information to provide weight data for subsequent clustering algorithms.

[0093] Since it is difficult to collect specific data on potential retailer customers, other indicators are not considered for the time being in this embodiment. Currently, only the customer type is used to represent the customer value for calculation. Based on the analysis of the existing real customers of this pharmaceutical logistics enterprise, the ratio of the number of each customer is obtained (hospital: health center: clinic: pharmacy: drugstore = 1:95:117:138:338). Also, since the comprehensive strength of this pharmaceutical logistics enterprise does not vary much for each customer, for simplicity of calculation, the average service price is based on the average freight rate for logistics from the central warehouse to other regions. Data such as the area, population, and GDP of each region in 2021 are obtained from the statistical yearbook, and the industry competition is analyzed based on the number of pharmaceutical wholesale enterprises in each region.

[0094] Classify the above indicators. The positive indicators are customer type, regional area, population, GDP, and the number of potential customers. The negative indicators are market share, the number of regional pharmaceutical logistics enterprises, and average freight rate.

[0095] Step 3: Use the entropy weight method to calculate the comprehensive evaluation result and perform normalization processing to obtain the conversion probability of potential customers.

[0096] S31: Process the values of the third-level indicators;

[0097] After constructing the evaluation index system for the conversion probability of potential pharmaceutical retailers, due to the large data gap, it is necessary to perform normalization processing on the third-level indicators. First, distinguish them into positive indicators and negative indicators according to the nature of the indicators.

[0098] The calculation formula for positive indicators is:

[0099]

[0100] The calculation formula for the negative index is as follows:

[0101]

[0102] where x ij represents the original value of the i-th customer under the j-th index, and n represents the number of all potential customers;

[0103] S32: Calculate the proportion P of the index value ij .

[0104]

[0105] where m represents the number of indices, and Y ij represents the standardized value after dimensionless processing of the i-th customer under the j-th index; P ij represents the proportion of the j-th index of the i-th customer in the sample value of this index, and is used as the probability when calculating the information entropy.

[0106] S33: Calculate the information entropy matrix of each potential pharmaceutical retailer customer. The calculation formula for the index entropy value is as follows:

[0107]

[0108] S34: Affect the weights of various indicators of the conversion probability of potential pharmaceutical retailers by calculating the information redundancy.

[0109] d j = 1 - E j , (j = 1, 2,..., m)

[0110]

[0111] where 1 - E j is the information entropy redundancy.

[0112] S35: Calculate the comprehensive score of each customer sample.

[0113]

[0114] S36: Perform normalization processing on the comprehensive score S i .

[0115] The specific role of normalization is to obtain a probability distribution between 0 and 1. The larger the original comprehensive score, the larger the result after normalization processing. That is to say, the higher the comprehensive score of potential pharmaceutical retailers, the greater the possibility that this customer will be converted into a real customer. The probability distribution after normalization processing is regarded as the probability that potential customers are converted into real customers, and this probability is used as the weight for considering potential customers in the subsequent clustering algorithm.

[0116] Step S4: Define the coverage area of the distribution center and screen out the customer points that can be delivered by the central warehouse.

[0117] Calculate the distance from each customer point to the warehouse to determine whether the customer point is directly delivered by the central warehouse or by the warehouse of the distribution center. The specific operation is as follows: According to the customer's geographical location, use the distance calculation tool in Excel to calculate the distance from the central warehouse to each pharmaceutical retailer customer point, sort them in ascending order of distance, take the warehouse as the center of the circle, and the distance from the warehouse to the farthest customer point as the radius. At this time, the distribution radius is set to 50 km. Screen out the customer data for which a distribution center needs to be established. In principle, the customer points within the circle are directly delivered by the warehouse, and the customer points outside the circle are delivered through the distribution center.

[0118] Step 5: Use the DBSCAN algorithm to remove remote points.

[0119] Remote points refer to a few clustering objects that are different from the normal data set. Since the K-means algorithm is sensitive to remote points, if the remote points are not removed in advance, it is very likely to affect the clustering effect. Therefore, this patent first calls the DBSCAN clustering algorithm encapsulated in Matlab, substitutes the preprocessed data, adjusts the two parameters of Eps and MinPts, and outputs the area where a distribution center needs to be established and the remote points (separately consign the customers of the remote points), and then performs K-means clustering on the data set after removing the remote points. In this embodiment, the coverage radius of the distribution center of the pharmaceutical logistics enterprise is set to 50 km. Since the DBSCAN clustering algorithm needs to independently determine the core object neighborhood radius Eps and the minimum density value Minpts, it is difficult to determine the optimal values. If the value of Eps is too small, too many clusters will be formed, and if the value is too large, all demand points will be clustered into one class. Therefore, the values of Eps and Minpts should be determined through multiple simulations in combination with the actual situation. Here, the neighborhood radius Eps is set to half of the coverage radius, that is, 25 km. According to geography, the distance corresponding to a 1-degree longitude difference in this area is about 89 km, and the distance corresponding to a 1-degree latitude difference is about 111 km. Therefore, the neighborhood radius Eps in coordinates is about 0.2, and the minimum density value MinPts is set to 25.

[0120] Step 6: Perform weighted K-means clustering with the potential customer conversion probability as the weight.

[0121] Perform weighted K-means clustering with the potential customer conversion probability as the weight. The weight of existing customers is 1, and potential customers are substituted into the calculation with the values after normalization in step S36. The calculation formula is:

[0122]

[0123] Write the weighted K-means code using Python for secondary clustering. Different from the traditional K-means algorithm, this patent takes into account the needs of potential customers, matches the longitude and latitude and weights for each pharmaceutical retailer customer, and sets the maximum weight for each cluster after clustering to constrain the size of the cluster area.

[0124] Algorithm input: Use the longitude, latitude and weights of the clustering objects as the data set, and set the maximum constraint.

[0125] Algorithm process:

[0126] (1) Initialization: Randomly select the smallest k objects as the initial clustering centers under the condition of meeting the constraints;

[0127] (2) Assign each clustering object to the most similar (nearest distance) cluster;

[0128] (3) Recalculate the minimum mean point in each cluster as the new cluster center;

[0129] (4) Calculate the silhouette coefficient to evaluate the clustering effect. If the clustering effect is poor, select a larger k value until the best k value for the clustering effect is selected;

[0130] (5) Repeat steps (2) and (3) until the centers of the k clusters no longer change.

[0131] Algorithm output: The best number of clusters and the clustering results.

[0132] The specific process diagram of the algorithm is as Figure 2 shown. The clustering result with weighted constraints obtained in this embodiment is as Figure 3 shown. Due to the limitation of the maximum constraint of each region, the algorithm clusters all customers into 6 categories with different coverage areas. The smaller the cluster area, the more customer demand in that area. Although the distribution ranges of each distribution center are different, the demand is relatively average. Therefore, this clustering result is relatively reasonable.

[0133] Step 7: Use the centroid method to find the location of the distribution center for each clustering area respectively.

[0134] First, determine the longitude and latitude of each retailer, mark them in the rectangular coordinate system, determine the Euclidean distance between each retailer, and find the centroid with the shortest sum of distances to all points in the region, which is the location point of the distribution center.

[0135] The centroid method has the following assumptions in the location calculation:

[0136] 1. The transportation cost is only related to the straight-line distance between the distribution center and the customer point;

[0137] 2. Do not consider the environmental factors and storefront prices of the location where the distribution center is located.

[0138] There are n pharmaceutical retailer customer points, distributed at different coordinate points (x j , y j ). Assuming that the distribution center is set at (x0, y0), the total transportation cost is:

[0139] where a j is the freight per unit weight and per unit distance from the distribution center to customer point j; w j is the transportation volume to customer point j; d j is the straight-line distance from the distribution center to customer point j:

[0140]

[0141] When selecting the distribution center, ensure that the total transportation cost is minimized.

[0142] The location selection results of the distribution centers in each region obtained by using the centroid method program are as Figure 4 shown.

[0143] In summary, the present invention provides a method for selecting the location of a distribution center considering potential pharmaceutical retailers. The description of the logistics distribution problem referred to in the present invention is as follows: With the continuous development and growth of end pharmaceutical logistics enterprises, new business scopes need to be opened up. The original central warehouse is not sufficient to support the operation of the entire enterprise, and the construction of the distribution center is urgent. A distribution center is a special type of distribution center. There has been a history of more than sixty years of research on distribution centers abroad, and research has only been carried out in China in recent years, but with remarkable results. Many scholars have developed various quantitative and qualitative methods to determine the location of distribution centers, which are also applicable to the location selection of distribution centers. Currently, the commonly used quantitative methods mainly include the centroid method, bilevel programming method, genetic algorithm, etc. Qualitative methods include the expert scoring method, Delphi method, etc.

[0144] Existing research has not considered the impact of the needs of potential customers on the location selection of distribution centers. However, in actual logistics work, logistics enterprises regard the needs of potential customers as an important indicator for the location selection of distribution centers. Facing a large-scale customer group, most scholars also adopt the method of clustering first and then location selection. When using the K-means algorithm to conduct clustering analysis on the longitude and latitude of customer points, potential customers and real customers are often treated equally. However, in real life, the contribution degrees of potential customers and real customers to clustering are not the same, and the probabilities of each potential customer converting into a real customer in the future are also different. Therefore, it is difficult to form a scientific and reasonable clustering result only by using the basic K-means algorithm.

[0145] As a further technical solution, when the present invention selects a location for a large-scale customer group, it considers the method of clustering first and then selecting a location. On the basis that the original K-means clustering method only considers the longitude and latitude of customer points, the present invention also considers the customer value of potential pharmaceutical retailers and uses this as a weight to design a K-means algorithm to achieve clustering under the maximum weight constraint, obtaining a more practical clustering result. Finally, the centroid method program is used to implement the location selection of the distribution center in each clustering area.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the purpose and scope of the present technical solution, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for locating a distribution center considering potential pharmaceutical retailers, characterized in that: It includes the following steps: S1: Obtain the existing customer order details of a pharmaceutical logistics enterprise and mine potential customer data; S2: Construct an evaluation index system for the conversion probability of potential pharmaceutical retailers; S3: Calculate the comprehensive evaluation result using the entropy weight method and perform normalization processing on it to obtain the conversion probability of potential customers; in step S3, it specifically includes the following steps: S31: Perform normalization processing on the values of the third-level indicators. First, distinguish them into positive indicators and negative indicators according to the nature of the indicators; The calculation formula for positive indicators is: The calculation formula for negative indicators is: where x ij represents the original value of the i-th customer under the j-th indicator, and n represents the number of all potential customers; S32: Calculate the proportion P of the index value ij : Among them, m represents the number of indicators, and Y ij represents the standardized value after dimensionless transformation of the i-th customer under the j-th indicator; P ij represents the proportion of the j-th indicator of the i-th customer in the sample value of this indicator, and this proportion is used as the probability when calculating the information entropy. S33: Calculate the information entropy matrix of each potential pharmaceutical retailer customer. The formula for calculating the index entropy value is: S34: Affect the weights of the indicators of the conversion probability of potential pharmaceutical retailers by calculating the information redundancy; d j = 1 - E j , (j = 1, 2, …, m) Among them, 1-E j is the information entropy redundancy; S35: Calculate the comprehensive scores of each customer sample; S36: Normalize the comprehensive score S i ; S4: Define the coverage range of the distribution center and screen out the customer points that can be delivered by the central warehouse; S5: Use the DBSCAN algorithm to remove remote points; S6: Perform weighted K-means clustering with the conversion probability of potential customers as the weight; in step S6, perform weighted K-means clustering with the conversion probability of potential customers as the weight. The weight of existing customers is 1, and the potential customers are calculated by substituting the values after normalization in step S36. The calculation formula is: Use the weighted K-means algorithm for secondary clustering, match the longitude and latitude and weight for each pharmaceutical retailer customer, and set the maximum weight for each cluster after clustering to constrain the size of the cluster area; S7: Use the centroid method to find the location of the distribution center for each clustering area respectively.

2. The method for locating a distribution center considering potential pharmaceutical retailers according to claim 1, wherein: In step S1, the existing customer information in the database of a pharmaceutical logistics enterprise includes: customer name customer, customer number customerID, customer address customerArea, customer longitude and latitude customerPosition. The process of mining potential pharmaceutical retailer data within the business radiation area of the enterprise using the descendant collector is as follows: S11: Select the target city served by the logistics enterprise in the map software; S12: Enter the key store names served by the logistics enterprise, including pharmacies, drugstores, clinics, health centers, and hospitals; S13: Collect store name, address, and phone information; S14: Save and export the collected data; S15: Compare the mined data with the existing customer data of the enterprise, screen out the list of potential customers to be developed, and convert all customer addresses into longitude and latitude.

3. The method for selecting the location of a distribution center considering potential pharmaceutical retailers according to claim 1, wherein: In step S2, use the customer value of potential pharmaceutical retailers, the comprehensive strength of pharmaceutical logistics enterprises, and the market role as secondary indicators, and then sort out the corresponding third-level indicators according to the secondary indicators to build an evaluation index system for the conversion probability of potential pharmaceutical retailers and determine the data that needs to be collected.

4. The method for allocating a distribution center considering potential pharmaceutical retailers according to claim 1, wherein: In step S4, when defining the coverage range of the distribution center and screening out the customer points that can be delivered by the central warehouse, it is necessary to divide the distribution ranges of the central warehouse and the distribution center according to the actual distribution requirements of the enterprise.

5. The method for allocating center site selection considering potential pharmaceutical retailers according to claim 1, characterized in that: In step S5, the preprocessed data is substituted into the DBSCAN clustering algorithm, and two parameters, Eps and MinPts, are adjusted to output the areas where distribution centers need to be established and outlying points.

6. The method for locating a distribution center considering potential pharmaceutical retailers according to claim 1, wherein: In step S7, the centroid method is used to find the locations of the distribution centers in each clustering area. First, the latitudes and longitudes of each retailer are determined and marked in a rectangular coordinate system, the Euclidean distances between the retailers are determined, and the centroid with the shortest sum of distances to all points within the area is found, which is the location point for the distribution center.

7. The method for selecting a distribution center location considering potential pharmaceutical retailers according to claim 6, wherein: The centroid method has the following assumptions in location calculation: (1) The transportation cost is only related to the straight-line distance between the distribution center and the customer points; (2) Environmental factors and storefront prices at the location of the distribution center are not considered; There are n pharmaceutical retailer customer points, distributed at different coordinate points (x j , y j ). Assuming that the distribution center is set at (x0, y0), the total transportation cost is: where a j is the freight per unit weight and unit distance from the distribution center to customer point j; w j is the transportation volume to customer point j; d j is the straight-line distance from the distribution center to customer point j: When selecting a distribution center, ensure that the total transportation cost is minimized.

Images