Opinion mining for thai restaurant reviews using K-Means clustering and MRF feature selection

Opinion mining on millions of Thai restaurant reviews in an unsupervised manner is a challenging task to survey feedbacks of the customers on their products and services. This is extremely helpful for owners to improve their business. In this paper, we propose an opinion mining on Thai restaurant reviews using K-Means clustering and MRF feature selection. The proposed method begins with text preprocessing for breaking reviews into words and removing stop words, followed by text transformation for creating keywords and generating input vectors. MRF feature selection is subsequently adopted for selecting relevant features from a large number of features extracted. Then, K-Means is employed for clustering into positive and negative reviews. From the experimental results, MRF feature selection can efficiently reduce the number of features in the data set so the computational time is significantly decreased. In addition, K-means can achieve the best clustering performance, when compared with Self-Organizing Map, Fuzzy C-Means, and Hierarchical Clustering. Thus, the cooperation of K-means with MRF feature selection is an effective model for clustering Thai restaurant reviews.