CUSTOMER SEGMENTATION AND CLASSIFICATION FROM BLOGS BY USING DATA MINING: AN EXAMPLE OF VOIP PHONE

Abstract

Blogs have been considered the 4th Internet application that can cause radical changes in the world, after e-mail, instant messaging, and Bulletin Board System (BBS). Many Internet users rely heavily on them to express their emotions and personal comments on whatever topics interest them. Nowadays, blogs have become the popular media and could be viewed as new marketing channels. Depending on the blog search engine, Technorati, we tracked about 94 million blogs in August 2007. It also reported that a whole new blog is created every 7.4 seconds and 275,000 blogs are updated daily. These figures can be used to illustrate the reason why more and more companies attempt to discover useful knowledge from this vast number of blogs for business purposes. Therefore, blog mining could be a new trend of web mining. The major objective of this study is to present a structure that includes unsupervised (self-organizing map) and supervised learning methods (back-propagation neural networks, decision tree, and support vector machines) for extracting knowledge from blogs, namely, a blog mining (BM) model. Moreover, a real case regarding VoIP (Voice over Internet Protocol) phone products is provided to demonstrate the effectiveness of the proposed method.

Keywords:

• Back-propagation neural network
• Blog
• Data mining
• Self-organizing map
• Sparse data
• Support vector machines

This work was supported in part by National Science Council of Taiwan (Grant No. NSC 96-2416-H-324-003-MY2).