The Research Of Customer Segmentation Based On Purchase Behavior

With the rapid development of the internet and the continuous improvement of the market,the global economy enters into the ear of e-commerce.Meanwhile,the difference between the goods and services provided by enterprises is gradully reduced.The method to purchase is more flexible.The choices is more diverse.The market competition is more intense.The enterprise's marketing strategy must change the center from production to customer,and change the target from sales to service.The mass marketing strategy must transfer to differentiated marketing strategy.Customers segmentation is the most important thing to establish and complete this kind of marketing system.There are two ways to customer segmentation.The first method is based on the customers' social attributes,consumption and other regular information.The second method is based on customers' purchase behavior.Customer segmentation based on customer behavior which also called customer segmentation based on retail transaction data is current research hotspot.It uses technologies of data mining to cluster customers.According to survey,current ways of customer segmentation based on purchase behavior is flawed and impractical.It only considers the products customer bought and lack of the category of them.However,many enterprises hava many kinds of product,so they divide products in small kinds,which lead to low group tightness.Some of the ways based on RFM model also ignored the differences in goods among the same group of customers.According to the existing problems in traditional customer segmentation based on purchase behavior,this paper puts forward a new customer clustering model,proposes a new customer clustering method based on this model and completes the following three innovative work:1.We propose a method that uses purchase tree to represent customer purchase information,which is a Hierarchical tree based on the category of the goods.The leaf node represents the goods and the internal node represents the goods category.When calculating the customer's distance,adding to the distance of internal category,it is convenient for us to analysis customer at different category level.Enhance the practical value of customer clustering results.2.Under the purchase tree model,this paper proposes a hierarchical clustering method.This method clustering the customer by purchase tree from high layer to low layer.Using the graph model in the same layer to improve the clustering result.3.In order to solve the problem exists in traditional grouping method which lacks consider of the goods' value.This paper proposes a method that introduces RFM value to purchase tree,create an RFM purchase tree for each customer,and then grouping the clusters.In this way,we can take the value of goods into consider.Using real data to do experiment,the results show that the new method has better customer clustering effect than the existing methods,and provides a new tool for the application of customer clustering.