Opinion classification using Maximum Entropy and K-Means Clustering

Abstract

Responds from academic questionnaire generally contains many comments, advice and suggestions. This responds is not processed systematically due to lack of method to process, whereas such information might be very useful as additional source in decision making. Opinion mining is well suited to address the issue. The objective of this study is to develop opinion classification system using Maximum entropy (ME) and K-Means Clustering (KMC). Opinion to be classified was Indonesian textual comments from academic questionnaire. Classification was conducted into two classes, i.e. negative opinion and positive opinion. Data contained of 2000 comments that was sampled as multi domain opinion, represented many objects such as lecturer, class room, etc. Features used for classification was selected from word in the opinion text. The weighting scheme that we used for clustering was TF/IDF. The results show that K-Means Clustering gives better performance as compared with ME in averages about 3% precision. KMC also perform faster than ME about 25 msec using 2000 text opinion.