KR102525076B1 - Method, Apparatus and System for Predicting Customer Defection Associated with Multiple Shopping Mall - Google Patents

Abstract

Disclosed are a customer churn prediction method, apparatus, and system that work with a plurality of shopping malls. A method for predicting customer churn linked with a plurality of shopping malls extracts customer order data corresponding to a first group from a plurality of shopping malls, and based on the extracted customer order data, a first customer churn prediction model corresponding to the first group and generating a second customer churn model corresponding to the first group by boosting the first customer churn prediction model using customer order data of the first shopping mall corresponding to the first group, and the learned second customer churn model Based on the prediction model, customer churn prediction information for predicting customer churn of the first shopping mall corresponding to the first group may be generated. Therefore, it is possible to control the shopping mall to predict customers' departure in advance and perform appropriate promotions by providing information specific to the shopping mall by utilizing big data.

Description

Method, Apparatus and System for Predicting Customer Defection Associated with Multiple Shopping Mall}

The present invention relates to a method, apparatus, and system for predicting customer churn that interworks with a plurality of shopping malls, and more particularly, analyzes behavior patterns of customers of a plurality of shopping malls to predict customer churn of shopping mall customers and utilize them for promotions. It relates to a method, apparatus, and system for predicting customer churn linked to a shopping mall.

In general, the popularization of online communication such as the Internet has radically changed the commercial environment more than ever. Now, e-commerce through online is so active that the word “online” or “internet shopping” itself is unfamiliar and feels like an unnecessary modifier. And the expansion trend of the online e-commerce market is expected to further expand.

BACKGROUND OF THE INVENTION [0002] Companies performing online e-commerce transactions, for example, shopping malls, are trying to attract new customers and prevent the departure of existing customers by performing marketing in various forms.

In particular, in the prior art, since the shopping mall cannot recognize the customers' departure from the shopping mall in advance, the customers' departure from the shopping mall cannot help but be neglected. Therefore, it is true that it is difficult to retain existing customers.

Korea Patent No. 1726404

The present invention is to solve this problem, and to provide a method, apparatus, and system for predicting customer churn linked with a plurality of shopping malls, which analyzes big data of shopping mall customers to predict customer churn and thereby utilize it for promotion. there is

In order to achieve this object, the present invention provides a customer churn prediction method that works with a plurality of shopping malls in one aspect (Aspect). The method for predicting customer churn is performed by a customer churn prediction device that interworks with a plurality of shopping malls, comprising: extracting customer order data corresponding to a first group from the plurality of shopping malls; generating a first customer churn prediction model corresponding to the first group based on the extracted customer order data; generating a second customer churn model corresponding to the first group by boosting the first customer churn prediction model using customer order data of the first shopping mall corresponding to the first group; and generating customer churn prediction information for predicting customer churn of the first shopping mall corresponding to the first group based on the learned second customer churn prediction model.

The extracting of the customer order data may include: collecting customer order data from the plurality of shopping malls; and extracting customer order data corresponding to the first group from the collected customer order data.

The customer order data includes, for each customer, within a set period, the number of orders for which a product has been ordered; the number of order cancellations that resulted in order cancellation of the product; the number of exchanges for which goods have been exchanged; number of orders placed in the last 1 unit period; number of orders placed in the last second unit period; number of orders placed in the last third unit period; the amount of orders placed during the last 1 unit period; order amount placed during the last second unit period; the amount of orders placed during the last third unit period; the number of logins during the last first unit period; the number of logins during the last second unit period; number of logins during the last 3 unit period; and a membership subscription period.

The customer order data may further include at least one of the number of mobile orders placed using a mobile device, the country of the customer placing the order, and the first inflow path through which the customer first introduced the product to purchase the product.

The customer churn prediction method may include, based on the generated customer churn prediction information, detecting information of a promotion target customer who needs a special promotion due to a high probability of churn among information of customers included in the first group; and transmitting the probability of departure for each customer and the information of the target customer for the promotion to the first shopping mall.

The step of detecting the information of the target customer for the promotion may include detecting information of a customer whose churn probability is greater than or equal to a set value based on the probability of churn for each customer calculated from the second customer churn prediction model as the information of the target customer for the promotion can include

The step of detecting the information of the target customer for the promotion may include calculating an expected loss for each customer based on the probability of churn for each customer calculated from the second customer churn prediction model and the sales for each customer extracted from the customer order data; and detecting information of a customer whose expected loss is greater than or equal to a set value based on the expected loss for each customer as the information of the target customer for the promotion.

On the other hand, in order to achieve the above object of the present invention, the present invention provides a customer departure prediction device that works with a plurality of shopping malls in another aspect. The customer churn prediction device may include an extraction module for extracting customer order data corresponding to a first group from a plurality of shopping malls; And based on the extracted customer order data, a first customer churn prediction model corresponding to the first group is generated, and the first customer churn using customer order data of the first shopping mall corresponding to the first group. By boosting the prediction model, a second customer churn model corresponding to the first group is generated, and based on the learned second customer churn prediction model, customer churn of the first shopping mall corresponding to the first group It may include an operation module for generating customer churn prediction information that predicts.

The extraction module may collect customer order data from the plurality of shopping malls, and extract customer order data corresponding to the first group from the collected customer order data.

The customer order data includes, for each customer, within a set period, the number of orders for which a product has been ordered; the number of order cancellations that resulted in order cancellation of the product; the number of exchanges for which goods have been exchanged; number of orders placed in the last 1 unit period; number of orders placed in the last second unit period; number of orders placed in the last third unit period; the amount of orders placed during the last 1 unit period; order amount placed during the last second unit period; the amount of orders placed during the last third unit period; the number of logins during the last first unit period; the number of logins during the last second unit period; number of logins during the last 3 unit period; and a member subscription period.

The customer order data may further include at least one of the number of mobile orders placed using a mobile device, the country of the customer placing the order, and the first inflow path through which the customer first introduced the product to purchase the product.

The operation module, based on the generated customer churn prediction information,

Among the information of customers included in the first group, it is possible to detect information of promotion target customers who are highly likely to churn and need a special promotion. The customer churn prediction device may further include a transmission unit that transmits the churn probability for each customer and the information of the target customer for the promotion to the first shopping mall.

The operation module may detect information of a customer whose churn probability is equal to or greater than a set value as information of the target customer for the promotion based on the churn probability for each customer calculated from the second customer churn prediction model.

The operation module calculates an expected loss for each customer based on the probability of churn for each customer calculated from the second customer churn prediction model and the sales for each customer extracted from the customer order data, and based on the expected loss for each customer Accordingly, information of a customer whose expected loss is greater than or equal to a set value may be detected as the information of the target customer for the promotion.

On the other hand, in order to achieve the above object of the present invention, the present invention provides a customer departure prediction system that works with a plurality of shopping malls in another aspect. The customer churn prediction system includes a plurality of shopping malls; and extracting customer order data corresponding to a first group from the plurality of shopping malls, generating a first customer churn prediction model corresponding to the first group based on the extracted customer order data, and By boosting the first customer churn prediction model using customer order data of the first shopping mall corresponding to , a second customer churn model corresponding to the first group is generated, and the learned second customer churn prediction model It may include a customer churn prediction device for generating customer churn prediction information for predicting customer churn of the first shopping mall corresponding to the first group based on.

The customer departure prediction device may collect customer order data from the plurality of shopping malls, and extract customer order data corresponding to the first group from the collected customer order data.

The customer order data includes, for each customer, within a set period, the number of orders for which a product has been ordered; the number of order cancellations that resulted in order cancellation of the product; the number of exchanges for which goods have been exchanged; number of orders placed in the last 1 unit period; number of orders placed in the last second unit period; number of orders placed in the last third unit period; the amount of orders placed during the last 1 unit period; order amount placed during the last second unit period; the amount of orders placed during the last third unit period; the number of logins during the last first unit period; the number of logins during the last second unit period; number of logins during the last 3 unit period; and a member subscription period.

The customer order data may further include at least one of the number of mobile orders placed using a mobile device, the country of the customer placing the order, and the first inflow path through which the customer first introduced the product to purchase the product.

The customer churn prediction device, based on the generated customer churn prediction information, among the information of the customers included in the first group, high possibility of churn, detects information of target customers who need a special promotion, and churn by customer Probability and information on the promotion target customer may be transmitted to the first shopping mall.

The customer churn prediction device may detect information of a customer whose churn probability is greater than or equal to a set value based on the churn probability for each customer calculated from the second customer churn prediction model as the information of the target customer for the promotion.

The customer churn prediction device calculates an expected loss for each customer based on the probability of churn for each customer calculated from the second customer churn prediction model and the sales for each customer extracted from the customer order data, and calculates the expected loss for each customer Based on this, information of a customer whose expected loss is greater than or equal to a set value may be detected as the information of the target customer for the promotion.

Meanwhile, in order to achieve the above object of the present invention, another aspect of the present invention provides a computer program stored in a medium. The computer program stored in the medium may include: extracting, by a computer, customer order data corresponding to a first group from a plurality of shopping malls; generating a first customer churn prediction model corresponding to the first group based on the extracted customer order data; generating a second customer churn model corresponding to the first group by boosting the first customer churn prediction model using customer order data of the first shopping mall corresponding to the first group; and generating customer churn prediction information for predicting customer churn of the first shopping mall corresponding to the first group based on the learned second customer churn prediction model.

As described above, according to the present invention, the departure of individual shopping mall customers is predicted based on the big data of a plurality of shopping mall customers, so that individual shopping malls can conduct promotions for potential customers based on information specific to the corresponding shopping mall. can

1 is a block diagram showing the configuration of a system for realizing a customer churn prediction method according to a preferred embodiment of the present invention.
Figure 2 is a flow chart showing the flow of the customer churn prediction method according to a preferred embodiment of the present invention.
3 is a conceptual diagram illustrating the concept of a customer churn prediction model generated and operated by a customer churn prediction device.
Figure 4 shows the results of predicting and verifying input variables included in customer order data for one year based on GBM for the next year.
5 and 6 show exemplary diagrams for explaining an example of an API service.
7 is a block diagram showing a system configuration for realizing a customer churn prediction method that works with a plurality of shopping malls.
8 is a flowchart for explaining a preferred embodiment of a method for predicting customer churn interworking with a plurality of shopping malls.
Figure 9 is an exemplary view for explaining the customer departure prediction method shown in Figure 8 by way of example.

Since the present invention can make various changes and have various embodiments, specific embodiments are illustrated in the drawings and described in detail. However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention.

Terms such as first and second may be used to describe various components, but the components should not be limited by the terms. These terms are only used for the purpose of distinguishing one component from another. For example, a first element may be termed a second element, and similarly, a second element may be termed a first element, without departing from the scope of the present invention. The terms and/or include a combination of a plurality of related recited groups or any group of a plurality of related recited groups.

It is understood that when a component is referred to as being 'connected' or 'connected' to another component, it may be directly connected or connected to the other component, but other components may exist in the middle. It should be. On the other hand, when a component is referred to as being 'directly connected' or 'directly connected' to another component, it should be understood that no other component exists in the middle.

Terms used in this application are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, terms such as 'include' or 'have' are intended to designate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, but one or more other features It should be understood that the presence or addition of numbers, steps, operations, components, parts, or combinations thereof is not precluded.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and unless explicitly defined in the present application, they should not be interpreted in an ideal or excessively formal meaning. don't

Hereinafter, with reference to the accompanying drawings, preferred embodiments of the present invention will be described in more detail. In order to facilitate overall understanding in the description of the present invention, the same reference numerals are used for the same components in the drawings, and redundant descriptions of the same components are omitted.

1 is a block diagram showing the configuration of a system for realizing a customer churn prediction method according to a preferred embodiment of the present invention, Figure 2 is a flow chart showing the flow of the customer churn prediction method according to a preferred embodiment of the present invention.

As shown in FIG. 1, the customer departure prediction device 100 may interwork with the shopping mall 10 and the database 20 through a communication network. The customer churn prediction device 100 is implemented by at least one computer and may be a server for performing a customer churn prediction method. The customer churn prediction device 100 may include a customer order data extraction module 110 , a customer churn prediction model operation module 120 and a transmission module 130 .

A customer may purchase a product from the shopping mall 10 through a series of processes of accessing the shopping mall 10, signing up as a member, ordering a product, and making payment. Order data related to customer orders may be stored and managed by the database 20 . The database 20 interworks with the shopping mall 10 and the customer churn prediction device 100. From an operational point of view, it may be included in the shopping mall 10 or may be included in a system for predicting customer churn. and may be included in third parties.

1 and 2, the customer order data extraction module 110 of the customer departure prediction device 100 may extract the customer order data of the first set period from the database 20 (step: S1). Here, the first set period may be arbitrarily determined according to the implementation environment, for example, 1 year, 2 years, etc. In this embodiment, the first set period is assumed to be 1 year from January 1, 2017 to December 31 I'm going to do it.

Customer churn prediction model operation module 120 of the customer churn prediction device 100 predicts customer churn in the second set period after the first set period based on the extracted customer order data of the first set period Customer churn It is possible to generate a customer churn prediction model that outputs predictive information (step: S2). For example, the customer churn prediction device 100 learns the customer churn probability using customer order data from January 1, 2017 to December 31, 2017, so that the second set period of one year in the future, for example, 2018 1 It is to create a customer churn prediction model that can predict customer churn from the 1st of the month to the 31st of December.

3 is a conceptual diagram for explaining the concept of a customer churn prediction model generated and operated by the customer churn prediction device 100 .

As shown in FIG. 3, the customer order data, which is an input variable, is the number of orders for each customer within a set period, the number of order cancellations for canceling product orders, the number of exchanges for exchanging products, and the last first unit Number of orders placed during the last unit period, Number of orders placed during the last 2nd unit period, Number of orders placed during the last 3rd unit period, Order amount placed during the last 1st unit period, Order amount placed during the last 2nd unit period, Last 3rd unit Amount of orders placed during the period, number of logins in the last 1st unit period, number of logins in the last 2nd unit period, number of logins in the last 3rd unit period, membership subscription period, number of mobile orders placed using mobile, country of the customer placing the order and an initial inflow path through which a customer first enters to purchase a product. In the embodiment shown in FIG. 3 , the first unit period may be the past year, the second unit period may be the past 90 days, and the third unit period may be the past 30 days.

The customer churn prediction model has a second period setting period as an output value, such as the number of purchases for the next year, the order amount for the next year, and the probability of churn during the next year (for example, the probability of not purchasing within the next year). , it can learn to output whether or not it deviated (TRUE/FALSE). Such information may be included in customer churn prediction information. For example, the customer churn prediction model may calculate a churn probability indicating a probability that a customer churns during a set period in the future, and the customer churn prediction information may include the churn probability.

Meanwhile, the customer order data extraction module 110 may extract customer order data of the second setting period. Customer churn prediction device 100 may generate actual customer churn information including information of customers who actually churn in the second set period based on the customer order data of the second set period (step: S3).

Next, the customer churn prediction device 100 corrects the customer churn prediction model based on verifying the customer churn prediction information predicted from the customer churn prediction model based on the actual customer churn information of the second set period It can (step: S4). For example, the customer churn prediction device 100 may correct the algorithm and values of the customer churn prediction model so that customer churn prediction information from January 1 to December 31, 2018 matches actual customer churn information. .

Next, the customer churn prediction model operation module 120 of the customer churn prediction device 100 inputs at least one of the customer order data of the first set period and the customer order data of the second set period into the calibrated customer churn prediction model By doing so, it is possible to generate customer churn prediction information for predicting customer churn of the third set period after the second set period (step: S5).

For example, the customer churn prediction device 100 is customer order data from January 1, 2017 to December 31, customer order data from January 1, 2018 to December 31, or two customer order data It is possible to generate customer churn prediction information that predicts customer churn from a third set period, for example, from January 1 to December 31, 2018, by inputting to the customer churn prediction model. Here, the customer churn prediction information may include the number of purchases for one year in 2019, order amount, churn probability, and whether or not to churn (TRUE/FALSE).

The customer churn prediction model operation module 120 of the customer churn prediction device 100 is based on the customer churn prediction information for predicting customer churn in the third set period, information on promotional target customers who are highly likely to churn and require special promotions. can be detected (step: S6).

For example, the customer churn prediction device 100 may detect information of a customer whose churn probability is greater than or equal to a set value as information on a target customer for promotion, based on the churn probability for each customer calculated from the customer churn prediction model.

Alternatively, the customer churn prediction device 100 calculates the expected loss for each customer based on the customer-specific churn probability calculated from the customer churn prediction model and the sales volume for each customer extracted from the customer order data, and based on the expected loss for each customer Accordingly, information of a customer whose expected loss is greater than or equal to a set value may be detected as the information of the target customer for the promotion.

The transmission module 130 may transmit the probability of departure for each customer and the information of the target customer for the promotion to the shopping mall 10 (step: S7). Then, the shopping mall 10 may use the transmitted information for various promotions such as discount events, coupon provision, and special reward points for promotion target customers.

As such, the method for predicting customer attrition according to a preferred embodiment of the present invention can predict customer attrition (eg, Customer Churn, dormant conversion, etc.) using past purchase patterns of each customer included in customer order data.

For example, as mentioned above, the input variables to the customer churn prediction model here include multiple types of indicators, such as the number of purchases and the number of logins during a specific previous set period of time, followed by churn that does not make a purchase during a specific set period of time. Customers can be predicted. The shopping mall 10 may conduct a promotion targeting customers who are expected to churn through customer churn prediction and have a high previous sales contribution. The logic of a customer churn prediction model that can be used for such learning may be GBM (Gradient Boost), GLM (General Linear Model), Deep Learning, Random Forest, and the like.

Figure 4 shows the results of predicting and verifying input variables included in customer order data for one year based on GBM for the next year. The orange color of the graph represents FALSE and the light blue color represents TRUE.

As shown in FIG. 4 , it can be seen that a member with a low predicted departure value rarely churns away, as in the left portion, and a member with a high churn prediction value actually leaves, as shown in the right portion. The customer churn prediction device 100 may further improve reliability based on verifying the customer churn prediction model using the predicted information and the actual customer churn information.

5 and 6 show exemplary diagrams for explaining an example of an API service.

As shown in FIG. 5, the web service may use Spring Boot. For example, you can use the "Minimal Spring Framework" packaged by Pivotal. The execution file is a single Jar file, H2o model file Spring MVC file (Spring Java file creation automation is possible) is used, and various Spring Plug-Ins are available. Execution is possible as an independent web service through the built-in Tomcat.

As shown in FIG. 6, as a single jar file, it can be distributed in various ways and can be used in the Spark/Hadoop environment. The learning time is relatively long, but the application (eg, prediction) time is very short. Learning data (eg, number of customers) is about 200,000 cases, and learning time may take about 2-3 minutes per model. It can be configured to enable real-time predictions using data pipelines.

On the other hand, the customer departure prediction device 100 may generate customer departure prediction information corresponding to a specific group based on the customer order data of the specific group. , Based on customer order data from multiple shopping malls, a basic customer churn prediction model corresponding to a specific group is built, and a basic customer churn prediction model is additionally trained based on customer order data from individual shopping malls to create a model specialized for each shopping mall for a specific group. An individual shopping mall application model may be established.

7 is a block diagram showing a system configuration for realizing a method for predicting customer churn linked with a plurality of shopping malls, and FIG. 8 is a flowchart for explaining a preferred embodiment of a method for predicting customer churn linked with a plurality of shopping malls. 9 is an exemplary diagram for explaining the method of predicting customer departure shown in FIG. 8 by way of example.

As shown in Figure 7, the customer departure prediction device 100 may work with a plurality of shopping malls (SM1, ..., SMn) and the database (20). 7 to 9, the customer order data extraction module 110 of the customer departure prediction device 100 extracts customer order data corresponding to a first group from a plurality of shopping malls (SM1, ..., SMn). It can (step: S11).

For example, the customer order data extraction module 110 of the customer departure prediction device 100 selects a first group from a plurality of shopping malls SM1, ..., SMn, that is, the first shopping mall SM1 to the nth shopping mall SMn. Customer order data of customers in their 20s can be extracted.

On the other hand, customer order data of the first shopping mall SM1 to nth shopping mall SMn may be stored in the database 20, and the customer order data extraction module 110 extracts the customer order data from the database 20. You may.

The customer churn prediction model operation module 120 of the customer churn prediction device 100 may generate a first customer churn prediction model corresponding to the first group based on the extracted customer order data (step: S12). The first customer churn prediction model may be a basic customer churn prediction model based on customer order data extracted from a plurality of shopping malls (SM1, ..., SMn) for a first group, for example, customers in their twenties.

Next, the predictive model operation module 120 of the customer departure prediction device 100 may extract customer order data of the first shopping mall SM1 corresponding to the first group from the extracted customer order data. In addition, the customer departure prediction device 100 corresponds to the first group by performing boosting to additionally learn the first customer departure prediction model using customer order data of the first shopping mall SM1 corresponding to the first group. It is possible to generate a second customer churn prediction model (step: S13).

The second customer churn prediction model is a dedicated model in which the customer churn characteristics of the first shopping mall SM1 are highlighted in the basic customer churn prediction model, and is a customer churn prediction model specialized for the first shopping mall SM1 corresponding to the first group. can

Subsequently, the customer departure prediction model operation module 120 of the customer departure prediction device 100 determines customer departure of the first shopping mall SM1 corresponding to the first group based on the learned second customer departure prediction model. It is possible to generate predictive customer churn prediction information (step: S14).

Subsequently, the customer churn prediction model operation module 120 of the customer churn prediction device 100 may detect information of a promotion target customer who is highly likely to churn and requires a special promotion based on the generated churn prediction information (step :S15).

For example, the customer churn prediction apparatus 100 may detect information of a customer whose churn probability is greater than or equal to a set value as information on a target customer for promotion based on a churn probability for each customer from a customer churn prediction model. Alternatively, the customer churn prediction device 100 calculates the expected loss for each customer based on the customer-specific churn probability calculated from the customer churn prediction model and the sales volume for each customer extracted from the customer order data, and based on the expected loss for each customer Accordingly, information of a customer whose expected loss is greater than or equal to a set value may be detected as the information of the target customer for the promotion.

The transmission module 130 may transmit the probability of departure for each customer and the information of the target customer for the promotion to the first shopping mall SM1 (step: S16). Then, the first shopping mall SM1 can use the transmitted information for various promotions such as discount events, coupon provision, and special reward points targeting the first group of promotion target customers.

Although the present invention has been described by exemplifying preferred embodiments, those skilled in the art will variously modify and change the present invention within the scope not departing from the technical details and scope of the present invention described in the claims below. You will understand that this can be done. Accordingly, changes to future embodiments of the present invention will not deviate from the technology of the present invention.

10: Shopping mall
20: database
100: Customer churn prediction device
110: Extract customer order data ??
120: Customer churn prediction model operation module
130: transmission module

Claims (22)

In the customer churn prediction method performed by a customer churn prediction device linked with a plurality of shopping malls,
extracting customer order data corresponding to a first group from the plurality of shopping malls;
generating a first customer churn prediction model corresponding to the first group based on the extracted customer order data;
generating a second customer churn prediction model corresponding to the first group by boosting the first customer churn prediction model using customer order data of the first shopping mall corresponding to the first group; and
Generating customer churn prediction information for predicting customer churn of the first shopping mall corresponding to the first group based on the learned second customer churn prediction model;
The first group is the first age group,
The second customer churn prediction model is a dedicated model in which the customer churn characteristics of the first shopping mall are highlighted in the first customer churn prediction model, which is a basic customer churn prediction model, and is specialized for the first shopping mall corresponding to the first group. Customer churn prediction method, characterized in that the customer churn prediction model.
The method of claim 1, wherein extracting the customer order data comprises:
collecting customer order data from the plurality of shopping malls; and
Customer churn prediction method comprising the step of extracting customer order data corresponding to the first group from the collected customer order data.
The method of claim 1, wherein the customer order data
Within the set period for each customer,
the number of orders for which the product was ordered;
the number of order cancellations that resulted in order cancellation of the product;
the number of exchanges for which goods have been exchanged;
number of orders placed in the last 1 unit period;
number of orders placed in the last second unit period;
number of orders placed in the last third unit period;
the amount of orders placed during the last 1 unit period;
the amount of orders placed during the last second unit period;
the amount of orders placed during the last third unit period;
the number of logins during the last first unit period;
the number of logins during the last second unit period;
number of logins in the last 3 unit period; and
Customer churn prediction method comprising at least one of the membership subscription period.
The method of claim 3, wherein the customer order data,
The method of predicting customer churn, characterized in that it further comprises at least one of the number of mobile orders placed using mobile, the country of the customer who placed the order, and the first inflow path through which the customer first entered to purchase the product.
The method of claim 1, based on the generated customer churn prediction information,
Detecting information of customers included in the first group who are highly likely to churn and need a special promotion; and
The method of predicting customer churn, characterized in that it further comprises the step of transmitting the probability of churn for each customer and the information of the target customer for the promotion to the first shopping mall.
The method of claim 5, wherein the step of detecting the information of the target customer for the promotion comprises:
Based on the churn probability for each customer calculated from the second customer churn prediction model, detecting information of a customer whose churn probability is greater than or equal to a set value as information of the target customer for the promotion. Customer churn prediction method.
The method of claim 5, wherein the step of detecting the information of the target customer for the promotion comprises:
Calculating an expected loss for each customer based on the probability of churn for each customer calculated from the second customer churn prediction model and the sales for each customer extracted from the customer order data; and
Customer churn prediction method comprising the step of detecting information of a customer whose expected loss is greater than or equal to a set value based on the expected loss for each customer as information of the target customer for the promotion.
an extraction module for extracting customer order data corresponding to a first group from a plurality of shopping malls; and
Based on the extracted customer order data, a first customer churn prediction model corresponding to the first group is generated, and the first customer churn is predicted using customer order data of the first shopping mall corresponding to the first group By boosting the model, a second customer churn prediction model corresponding to the first group is generated, and based on the learned second customer churn prediction model, customer churn of the first shopping mall corresponding to the first group Includes an operation module for generating customer churn prediction information that predicts,
The first group is the first age group,
The second customer churn prediction model is a dedicated model in which the customer churn characteristics of the first shopping mall are highlighted in the first customer churn prediction model, which is a basic customer churn prediction model, and is specialized for the first shopping mall corresponding to the first group. Customer churn prediction device, characterized in that the customer churn prediction model.
The method of claim 8, wherein the extraction module,
Customer churn prediction apparatus, characterized in that for collecting customer order data from the plurality of shopping malls, and extracting customer order data corresponding to the first group from the collected customer order data.
9. The method of claim 8, wherein the customer order data is
Within the set period for each customer,
the number of orders for which the product was ordered;
the number of order cancellations that resulted in order cancellation of the product;
the number of exchanges for which goods have been exchanged;
number of orders placed in the last 1 unit period;
number of orders placed in the last second unit period;
number of orders placed in the last third unit period;
the amount of orders placed during the last 1 unit period;
order amount placed during the last second unit period;
the amount of orders placed during the last third unit period;
the number of logins during the last first unit period;
the number of logins during the last second unit period;
number of logins during the last 3 unit period; and
Customer churn prediction device comprising at least one of the membership subscription period.
The method of claim 10, wherein the customer order data,
Customer churn prediction device further comprising at least one of the number of mobile orders placed using a mobile, the customer's country and the customer's first inflow path to purchase a product.
The method of claim 8, wherein the operation module,
Based on the generated customer churn prediction information,
Among the information of customers included in the first group, information of customers who are highly likely to churn and need special promotions is detected,
Customer churn prediction device further comprising a transmission unit for transmitting the probability of churn for each customer and the information of the target customer for the promotion to the first shopping mall.
The method of claim 12, wherein the operation module,
Customer churn prediction device, characterized in that for detecting information of a customer whose churn probability is greater than a set value as information of the target customer based on the churn probability for each customer calculated from the second customer churn prediction model.
The method of claim 12, wherein the operation module,
Estimated loss for each customer is calculated based on the probability of churn for each customer calculated from the second customer churn prediction model and sales for each customer extracted from the customer order data, and the expected loss is set based on the expected loss for each customer Customer churn prediction device, characterized in that for detecting the information of the customer having a value greater than or equal to the information of the target customer for the promotion.
delete delete delete delete delete delete delete delete

Images