A method and device for selecting retail customer demand information samples based on customer domain

Abstract

The present invention relates to a method and device for selecting samples of retail customer demand information based on customer domains, and belongs to the technical field of big data mining and customer classification management. The present invention is based on the data characteristics of two dimensions, namely, customer business characteristics and market characteristics, to perform homogeneous division of retail customers, and generate retail customer domains with common business characteristics and consumer groups; select specimens through the continuity characteristics, stability characteristics, and integrity management characteristics of operations; compositely calculate the sample comprehensive index based on the average monthly purchase volume, average monthly single value, and brand width as the characteristics, and select samples in proportion with the mean of the comprehensive index of samples of members in the domain as the center; verify the coverage of samples through historical data, recursively check, and select the optimal samples. The sample selection method of the present invention can take into account the homogeneity and representativeness of samples, ensure the coverage and comprehensiveness of the screened samples, and provide an accurate data source for subsequent data analysis.

Description

Technical Field

[0001] The present invention relates to a method and device for selecting retail customer demand information samples based on customer domain, belonging to the technical field of big data mining and customer classification management. Background Art

[0002] With the rapid development of science and technology, the continuous expansion and popularization of mobile networks, the wave of informatization has quickly swept the world. At the same time, a huge amount of stored data has appeared in all walks of life. While people realize the importance of information, they have higher and higher requirements for information analysis and use. How to use effective methods in big data to mine deep and useful information behind the data has become the focus and focus of related industries. At present, a series of data mining related technologies and industries have emerged and are developing rapidly.

[0003] For enterprises in the retail industry, such as tobacco commercial enterprises, in-depth mining of market demand information is very important. On the one hand, enterprises can effectively organize supply sources and adjust delivery strategies based on market demand information to achieve accurate delivery, effectively meet customer needs and improve the quality of enterprise economic operation; on the other hand, it can help grassroots marketing personnel analyze customer potential needs, provide differentiated services in a targeted manner, effectively meet customers' personalized needs, further enhance customer profitability, improve customer relationships, enhance customer stickiness, and have important significance for the extension of the marketing chain. At present, the in-depth mining of market demand information is based on the demand information of retail customers. In the case of a large number of samples of retail customer demand information, how to effectively select samples is very important. At present, most of the sample selection is based on the probability distribution of the sample, without considering the market characteristics and business characteristics of the retail customers themselves, and the coverage rate of the selected samples is not high. Summary of the invention

[0004] The purpose of the present invention is to provide a method for selecting retail customer demand information samples based on customer domain, so as to solve the problems that the current sample selection method does not take customer market characteristics and business characteristics into consideration and has low sample coverage.

[0005] In order to solve the above technical problems, the present invention provides a method for selecting retail customer demand information samples based on customer domain, and the selection method comprises the following steps:

[0006] 1) Obtain a sample of retail customer demand information in the research area, which includes basic characteristics and business characteristics of retail customers, where business characteristics include average monthly sales and single item value, basic characteristics include market characteristics, and market characteristics include market type, business format and business district;

[0007] 2) According to the average monthly sales volume, retail customers are divided into x categories, and according to the single value, retail customers are divided into y categories, thereby dividing retail customers into xy initial domains. On this basis, cross-division is performed according to market type, business format and business district, and the initial domain is divided into multiple retail customer domains; customers in each retail customer domain have the same category of average monthly sales volume, the same category of single value, the same market type, the same business district and the same business format;

[0008] 3) screening the retail customers in each retail customer domain based on at least one of the characteristics of business continuity, stability and honest business, to obtain valid retail customer data in each retail customer domain;

[0009] 4) In each retail customer domain, the sample comprehensive value is calculated with the average monthly purchase volume, the average monthly single item value and the brand width as the characteristics, the mean of the sample comprehensive index in each retail customer domain is calculated, and the valid retail customer data sample closest to the center in the corresponding retail customer domain is selected according to the set ratio with the mean of the sample comprehensive index in each retail customer domain as the center; wherein the brand width refers to the number of commodity specifications contained in the historical purchase data of the retail customer in the previous n months; for a retail customer domain with a small number of customers, the set ratio is set relatively high, and for a retail customer domain with a large number of customers, the set ratio is set relatively low;

[0010] The selected samples shall meet the verification conditions: the ratio of customers represented by the domain covered by the selected samples to the total number of customers is greater than the set value, and the ratio of the average monthly purchase share of the domain members covered by the selected samples to the monthly purchase volume of all customers is greater than the set value; if the verification conditions are not met, the set ratio shall be adjusted until the verification conditions are met.

[0011] The present invention also provides a device for selecting retail customer demand information samples based on customer domain, the device comprising: a processor; a memory for storing executable instructions of the processor; wherein the processor is configured to execute the method for selecting retail customer demand information samples based on customer domain described in the present invention by executing the executable instructions.

[0012] The present invention divides the retail customer domain with common business characteristics and consumer groups according to business characteristics and market characteristics, and ensures the homogeneity of the business characteristics and consumer group characteristics of the retail customer domain; at the same time, the continuity characteristics, the stability characteristics and the integrity management characteristics are used to screen the retail customers, and the average monthly purchase volume, the average monthly single value and the brand width are used as the characteristics to calculate the comprehensive value of the sample, and ensure the representativeness of the selected sample. Therefore, the sample selection method of the present invention can take into account the homogeneity and representativeness of the sample, ensure the coverage and comprehensiveness of the sample screened out, and provide an accurate data source for subsequent data analysis.

[0013] Furthermore, in order to ensure that the generated retail customer domain can fully cover the business characteristics and market characteristics, the division process of step 2) is as follows: segmentation is performed according to two indicators in the business characteristics to obtain the initial domain; the initial domain is cross-divided by the three indicators of market type, business format and business district to obtain retail customer domains with common business characteristics and consumer groups.

[0014] Furthermore, in order to achieve accurate classification of business characteristics, the market type is divided into urban and rural categories, the business format is divided into grocery stores, tobacco and alcohol convenience stores, and others, and the business district is divided into residential areas, business and government districts, and others.

[0015] Furthermore, the continuity feature in step 3) refers to the retail customer's continuous operation exceeding the set time, and the verification standard is to trace back from the current time to the set time, and the monthly purchase volume is not 0; the stability feature refers to the retail customer's operating stability index is less than or equal to the set threshold; the integrity management feature is that the retail customer's integrity management level meets the set requirements.

[0016] Furthermore, in order to ensure that the screened retail customer data is valid, the valid retail customer data refers to samples that simultaneously meet the requirements of continuity, stability, and integrity management.

[0017] Furthermore, in order to accurately calculate the sample comprehensive index, the calculation process of the sample comprehensive index in step 4) is as follows:

[0018] Normalize the average monthly purchase volume, average monthly single item value, and brand breadth of each retail customer in each customer domain to obtain the corresponding average monthly purchase volume index, average monthly single item value index, and brand breadth index;

[0019] The weighted sum of the average monthly purchase volume index, average monthly single item value index, and brand breadth index is the sample comprehensive index of the retail customer. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flow chart of the method for selecting retail customer demand information samples based on customer domain of the present invention. DETAILED DESCRIPTION

[0021] The specific implementation of the present invention will be further described below in conjunction with the accompanying drawings.

[0022] Method Embodiment

[0023] The present invention divides retail customers into homogeneous groups based on the data features of customer business characteristics and market characteristics, and generates a retail customer domain with common business characteristics and consumer groups; selects specimens based on the continuity, stability, and integrity of operations; calculates the sample comprehensive index based on the average monthly purchase volume, average monthly single value, and brand breadth as the characteristics, and selects samples in proportion with the mean of the comprehensive index of the samples of members in the domain as the center; verifies the sample coverage through historical data, recursively checks, and selects the best sample. The implementation process of this method is as follows: Figure 1 As shown, the specific implementation process.

[0024] 1. Obtain a sample of demand information from retail customers in the study area.

[0025] The samples obtained by the present invention include basic characteristics and operating characteristics of retail customers. Operating characteristics include average monthly sales and single item value. Basic characteristics include market characteristics. Market characteristics include market type, business format and business district. Average monthly sales, single item value), the selected time span is the complete historical data of n months from the current time, and the single item value is calculated at the wholesale price (customer purchase price). The formula is as follows: single item value = purchase amount (yuan) in n months / total purchase volume (items) in n months. The size of n can be set according to actual conditions, for example, n can be 12; market types are: urban and rural areas, business formats are grocery stores, tobacco and alcohol convenience stores, and other three categories, and business districts are residential areas, business and government affairs areas, and other three categories.

[0026] 2. To homogenize the demand information samples of each retail customer in the study area and generate retail customer domains with common business characteristics and consumer groups.

[0027] Use the K-means algorithm to divide retail customers into x categories according to average monthly sales, and use the K-means algorithm to divide retail customers into y categories according to single value, and divide retail customers into xy initial domains; on this basis, cross-division is performed according to market type, business type and business district type, and the initial domain is divided into multiple retail customer domains, and customers in each customer domain have the same category of average monthly sales, the same category of single value, the same market type, the same business district and the same business type. As other implementation methods, the K-means algorithm here can also use other clustering algorithms.

[0028] 3. Filter the retail customers in each retail customer domain to obtain valid retail customer data in each retail customer domain.

[0029] The present invention selects effective samples through the continuity characteristics, stability characteristics, and honest operation characteristics of operations. The continuity characteristic refers to the continuous operation of the cigarette with zero customers for more than a months (a≤n), and the verification standard is to trace back a months from the current time, and the purchase volume of each month is not 0; the stability characteristic refers to the customer's business stability index is less than or equal to b, and the honest operation characteristics are (AAA, AA, A, B, C, D) B or above (including B); effective specimens refer to specimens that meet the requirements of continuity characteristics, stability characteristics, and honest operation characteristics at the same time. The calculation method of the stability characteristic index is: the discreteness of the ratio of the customer's historical purchase data in the previous n months to the monthly purchase total is magnified by c times. This embodiment uses K value segment clustering to determine that the stability characteristic index is less than or equal to b to meet the stability characteristic requirements. When selecting effective samples, you can also screen according to any one or two of the continuity characteristics, stability characteristics, and honest operation characteristics of operations.

[0030] 4. Taking the mean of the comprehensive index of samples in each retail customer domain as the center, select valid retail customer data samples in the corresponding retail customer domain according to the set ratio.

[0031] The present invention calculates the comprehensive value of samples in each retail customer domain based on the average monthly purchase volume, the average monthly single item value and the brand width as characteristics, calculates the mean of the comprehensive index of samples in each retail customer domain, and selects valid retail customer data samples in the corresponding retail customer domain according to a set ratio with the mean of the comprehensive index of samples in each retail customer domain as the center.

[0032] The brand width refers to the number of commodity specifications included in the historical purchase data of retail customers in the previous n months. When calculating the sample comprehensive value, the monthly average purchase volume, the monthly average single value, and the brand width are normalized to obtain the monthly average purchase volume index, the monthly average single value index, and the brand width index. The monthly average purchase volume, the monthly average single value, and the brand width are weighted. According to the monthly average purchase volume index p1%, the monthly average single value index p2%, and the brand width index p3%, the weights are assigned (p1+p2+p3=100). The weighted sum of the three indexes is the sample comprehensive index. The size of the weight of each index can be set according to actual needs. If the monthly purchase volume index is considered to be more important, its weight can be set larger. Through the above weighted summation, the sample comprehensive index of each retail customer can be calculated. For each retail customer domain, there are multiple effective customers screened out. In order to select effective customers in each retail customer domain, it is necessary to calculate the sample comprehensive index of all effective customers in each retail customer domain and calculate the mean. The mean of the sample comprehensive index in the retail customer domain is the center, and the customers closest to the center are selected in proportion. The proportion can be set according to experience. For a small number of customers in the retail customer domain, the selection ratio is relatively high, and for a large number of customers in the retail customer domain, the selection ratio is relatively low. For example, it can be selected between 3% and 15%. The sample selection ratio is calculated according to the number of customers who have not been screened as effective customers.

[0033] In order to ensure the coverage of the selected samples, the present invention recursively verifies the sample coverage through historical data. The sample coverage refers to the ratio of the customers represented by the domain covered by the sample to the total number of customers is greater than p4% to be qualified, and the ratio of the average monthly purchase share of the domain members covered by the sample to the average monthly purchase volume of all customers is greater than p4% to be qualified; if the recursive verification means that the sample coverage does not meet the requirements, the sample selection ratio is fine-tuned and the sample coverage verification is repeated until the sample coverage is qualified.

[0034] In order to better illustrate the effect of the present invention, the process of the present invention is described by taking cigarette retail customers as an example. Figure 1As shown in the figure, taking the sample selection of regional retail customer demand information collection in a certain city as an example, the basic data and business data of 18,016 cigarette retail customers in a certain city in China from May 2020 to April 2021 were selected, including two indicators of customer business characteristics, namely, average monthly sales and single-line value. The selected time span is the complete historical data of the current 12 months. The single-line value is calculated at the wholesale price (customer purchase price). The formula is as follows: single-line value = 12-month purchase amount (yuan) / 12-month purchase total (line). The cigarette retail customers are divided into homogeneous groups to generate cigarette retail customer domains with common business characteristics and consumer groups. The division method is: using the K-means algorithm to divide customers into 10 segments according to average monthly sales, using the K-means algorithm to divide customers into 7 segments according to single-line value, and dividing customers into 70 domains, see Table 1 for details.

[0035] Table 1

[0036]

[0037] According to 2 types of market, 3 types of business formats and 3 types of business districts, on the basis of the above 70 domains, the three indicators are crossed to divide customers into 1,260 domains, as shown in Table 2 for details; among them, there are 618 domains with customers.

[0038] Table 2

[0039]

[0040]

[0041] Valid samples are selected based on the continuity, stability and honest operation characteristics of the operation. The continuity characteristic refers to the cigarette being operated for more than 8 consecutive months without any customers. The verification standard is that the purchase volume of each month is not 0 when looking back 8 months from the current time. The stability characteristic refers to the customer's operation stability index being less than or equal to 0.5. The honest operation characteristic is (AAA, AA, A, B, C, D) Class B or above (including Class B). Valid samples refer to samples that meet the requirements of continuity, stability and honest operation characteristics at the same time.

[0042] The stability feature refers to the customer's business stability index less than or equal to 0.5. The calculation method of the stability feature index is: the dispersion of the ratio of the customer's historical purchase data in the previous 12 months to the monthly purchase volume is magnified 10,000 times. This model uses K value 9-segment clustering to determine that the stability feature index is less than or equal to 0.5 to meet the stability feature requirements;

[0043] The sample comprehensive index is calculated by weighting the average monthly purchase volume, average monthly single-line value, and brand breadth. The sample is selected in proportion with the mean of the sample index of members in the domain as the center. Brand breadth refers to the number of cigarette specifications included in the customer's historical purchase data in the previous 12 months.

[0044] Weighted composite calculation sample index: After normalizing the average monthly purchase volume, average monthly single value, and brand width, the average monthly purchase volume index, average monthly single value index, and brand width index are obtained. The weights are assigned according to the average monthly purchase volume index of 20%, the average monthly single value index of 45%, and the brand width index of 35%. The weighted sum of the three indexes is the sample comprehensive index. Select samples by proportion: Select samples by proportion according to the number of members in the domain. This region selected 1,342 samples based on the above method. The sample selection ratio is shown in Table 3 below:

[0045] Table 3

[0046]

[0047] The sample coverage is verified through historical data and recursively checked. In this example, the sample coverage refers to the ratio of customers represented by the domain covered by the sample to the total number of customers is greater than 95% to be qualified, and the ratio of the average monthly purchase share of domain members covered by the sample to the average monthly purchase volume of all customers is greater than 95% to be qualified. If the sample coverage does not meet the requirements, the sample selection ratio is fine-tuned and the sample coverage verification is repeated until the sample coverage is qualified. Through recursive verification, this example finally selected 1,342 samples, covering 17,600 customers, and the customer coverage rate reached 96.58%. The total sales of domain customers covered by the sample accounted for 97.73% of the total sales in the region.

[0048] Device Embodiment

[0049] The device for selecting retail customer demand information samples based on customer domain of the present invention comprises a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the method of the present invention for implementing the above method embodiment. That is, the method in the above method embodiment should be understood as a flow of the method for selecting retail customer demand information samples based on customer domain implemented by computer program instructions. These computer program instructions can be provided to the processor so that the processor executes these instructions to generate functions specified by the above method flow.

[0050] The processor referred to in this embodiment refers to a processing device such as a microprocessor MCU or a programmable logic device FPGA; the memory referred to in this embodiment includes a physical device for storing information, which usually digitizes the information and then stores it in a medium using electrical, magnetic or optical methods. For example: various memories that use electrical energy to store information, such as RAM, ROM, etc.; various memories that use magnetic energy to store information, such as hard disks, floppy disks, magnetic tapes, magnetic core memories, magnetic bubble memories, and U disks; various memories that use optical methods to store information, such as CDs or DVDs. Of course, there are other types of memories, such as quantum memories, graphene memories, and so on.

[0051] The device is composed of the above-mentioned memory, processor and computer program, and is implemented by the processor executing corresponding program instructions in the computer. The processor can be equipped with various operating systems, such as Windows operating system, Linux system, Android, iOS system, etc.

[0052] As another embodiment, the device may further include a display, and the display is used to display the selection results for reference by the staff.

[0053] The above is only the preferred embodiment of the present invention, and is not intended to limit the present invention. Many changes and modifications may be made without departing from the scope of the present invention. Therefore, it is intended that the above detailed description is considered to be illustrative rather than restrictive, and it should be understood that the following claims (including all equivalents) are intended to limit the spirit and scope of the present invention. The above embodiments should be understood to be only used to illustrate the present invention and not to limit the scope of protection of the present invention. After reading the content of the record of the present invention, the technician can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall into the scope defined by the claims of the present invention.

Claims

1. A method for selecting retail customer demand information samples based on customer domain, characterized in that: The selection method includes the following steps: 1) Obtain a sample of retail customer demand information in the research area. The sample includes basic characteristics and operating characteristics of retail customers. Operating characteristics include average monthly sales and single item value. Basic characteristics include market characteristics. Market characteristics include market type, business format and business district. 2) Retail customers are divided into x categories according to their average monthly sales, and into y categories according to their single item values, thereby dividing retail customers into xy initial domains. On this basis, cross-division is performed according to market type, business format, and business district, and the initial domain is divided into multiple retail customer domains; customers in each retail customer domain have the same average monthly sales of the same category, the same single item value of the same category, the same market type, the same business district, and the same business format; 3) Screen the retail customers in each retail customer domain based on the characteristics of business continuity, stability and integrity, and obtain valid retail customer data in each retail customer domain; 4) In each retail customer domain, the sample comprehensive value is calculated based on the average monthly purchase volume, the average monthly single item value and the brand width as the characteristics, and the mean of the sample comprehensive index in each retail customer domain is calculated. The valid retail customer data sample closest to the center in the corresponding customer domain is selected according to the set ratio with the mean of the sample comprehensive index in each retail customer domain as the center; the brand width refers to the number of commodity specifications included in the historical purchase data of the retail customer in the previous n months; for retail customer domains with a small number of customers, the set ratio is set relatively high, and for retail customer domains with a large number of customers, the set ratio is set relatively low; The selected samples shall meet the verification conditions: the ratio of customers represented by the domain covered by the selected samples to the total number of customers is greater than the set value, and the ratio of the average monthly purchase share of the domain members covered by the selected samples to the monthly purchase volume of all customers is greater than the set value; if the verification conditions are not met, the set ratio shall be adjusted until the verification conditions are met.

2. The method for selecting retail customer demand information samples based on customer domain according to claim 1, characterized in that: The division process of step 2) is as follows: segment and divide according to two indicators in the business characteristics to obtain an initial domain; cross-divide the initial domain by three indicators of market type, business format and business district to obtain a retail customer domain with common business characteristics and consumer groups.

3. The method for selecting retail customer demand information samples based on customer domain according to claim 2, characterized in that: According to market type, it is divided into two categories: urban and rural areas; according to business format, it is divided into three categories: grocery stores, tobacco and alcohol convenience stores, and others; according to business district, it is divided into three categories: residential areas, business and government districts, and others.

4. The method for selecting retail customer demand information samples based on customer domain according to claim 1, characterized in that: The continuity feature in step 3) means that the retail customer has been operating continuously for more than the set time, and the verification standard is to trace back from the current time to the set time, and the monthly purchase volume is not 0; the stability feature means that the retail customer's operating stability index is less than or equal to the set threshold; the integrity management feature means that the retail customer's integrity management level meets the set requirements.

5. The method for selecting retail customer demand information samples based on customer domain according to claim 1 or 4, characterized in that: The valid retail customer data mentioned above refers to samples that meet the requirements of continuity, stability, and integrity management.

6. The method for selecting retail customer demand information samples based on customer domain according to claim 1, characterized in that: The calculation process of the sample comprehensive index in step 4) is as follows: Normalize the average monthly purchase volume, average monthly single item value, and brand breadth of each retail customer in each retail customer domain to obtain the corresponding average monthly purchase volume index, average monthly single item value index, and brand breadth index; The weighted sum of the average monthly purchase volume index, average monthly single item value index, and brand breadth index is the sample comprehensive index of the retail customer.

7. A retail customer demand information sample selection device based on customer domain, characterized in that: The device comprises: a processor; a memory for storing executable instructions of the processor; wherein the processor is configured to execute the method for selecting retail customer demand information collection samples based on customer domain according to any one of claims 1 to 6 by executing the executable instructions.

Images