英语论文网

Abstract: Understanding and adapting to changes in customerbehavior is an important aspect for survival in a continuouslychanging environment.
This paper develops a methodology basedon decision tree analysis to detect the change in classified customersegments automatically between two data sets collected over time.We first define three types of changes as the emerging pattern, theunexpected change and the added=perished rule. Then, similarityand difference measures are developed for rule matching to detectall types of change. Finally, the degree of change is developed toevaluate the amount of change. Our suggested methodology basedon decision tree analysis in the change detection problem can beused in more structured situations in which the manager has aspecific research question and it also detects the change ofclassification criteria in a dynamically changing environment. AKorean Internet shopping mall case is evaluated to represent theperformance of our suggested methodology, and practical businessimplications for this methodology are also provided. We believethat the change detection problem and the suggested methodologywill become increasingly important as more data mining applicationsare implemented.

Keywords: data mining, decision tree, change analysis,Internet shopping

mall1. IntroductionUnderstanding and adapting to changes in customer needsis an important aspect for survival in a continuouslychanging environment. In this paper, we propose amethodology for mining the change in customer behaviorbefore and after a certain point in the contexts of decisiontree classification.More specifically, this paper aims to giveanswers about the following questions. What are thedifferences in customer segmentation before and after anew service adoption?Which customer group’s purchasingamount is gradually increasing? Which customer groupsare churned over the years? These are typical questionswhich relate to the change in customer classification.Recent results of interviews performed by a consultingcompany show the importance of knowing the change incustomer needs.For this purpose, we discover the change in customerbehavior by comparing two classification rule sets whichare generated from different periods. Our suggestedmethodology is based on data mining techniques, especiallydecision tree analysis. Data mining is the process ofexploration and analysis of large quantities of data inorder to discover meaningful patterns and rules. In thiscontext, data mining shows great potential for trackingchanges in customer behavior automatically. Therefore, weuse data mining techniques to detect changes in customerbehavior. But most data mining techniques such asassociation rule mining, decision trees and neural networkscannot be applied alone to answer the above researchquestions, because they cannot handle dynamic situationswell.Liu et al. (2000) suggest a change mining methodologybased on decision trees to find interesting relationshipsamong a large set of data items. However, their researchcannot give an answer about how much change hasoccurred. It is important to detect what kinds of changeshave occurred, but it is also important to differentiatewhich changes are more serious. Song e本论文由英语论文网提供整理，提供论文代写，英语论文代写，代写论文，代写英语论文，代写留学生论文，代写英文论文，留学生论文代写相关核心关键词搜索。