An Intelligent Transformation Knowledge Mining Method Based on Extenics
Xingsen Li, Haolan Zhang, Zhengxiang Zhu, Zhongbiao Xiang, Zhengxin Chen, Yong Shi · 網際網路技術學刊 · 2013
With the rapid development of internet technology, more and more businesses are running on the Web and lots of business data have been stored in databases or log files in web servers. How to make better decisions on the huge amount of data is an urgent task. Knowledge acquisition through data mining becomes one of the most important directions to support scientific decision making; however, the knowledge discovered from data mining may not work effectively because most of it describes only static knowledge, not how-to-do knowledge. In this paper, we describe our approach of dealing with this problem, starting with two existing methods: MCLP (Multiple Criteria Linear Programming)-based method classifies the web data sets and find the best match elements, but the models obtained are unexplainable; decision tree methods can obtain explainable rules, but they are static know-what knowledge-for example, we still don't know how to transfer from class bad to class good. In order to convert from such passive result to actionable knowledge, we focus on a new methodology for discovering actionable knowhow knowledge, which is based on decision tree rules and Extension set theory. As a second order mining method, our approach can extract a set of hidden patterns describing how to change churn customers' behaviors and provide ”from can't to can, from bad to good” rules. It can provide successful decision making support on the transformation of the customer churn. A case study of a web company has shown that our method is feasible and effective. In addition, the case study also has the reference value for data mining in other web businesses.