Automatic Thai Folk Wisdom Classification using Data Mining Approach

Abstract

This research aims to propose the automatic Thai folk wisdom classification using data mining approaches from text description. The process consists of 4 steps, 1) word extraction and stop words elimination 2) feature selection 3) classification of model construction and 4) model evaluation. In order to compare the accuracy results of Decision Tree, K-Nearest Neighbor,
Distance-Weight K-Nearest Neighbor and Naïve Bayes, K-folds cross validation technique over 500 Thai folk wisdom data were used. The comparison result of this study showed that the highest performance is Naïve Bayes with accuracy 93.5%. As this result, we selected the Naïve Bayes classifier for implementing a web-based application using WEKA API and PHP. This
tool supports the ease of use when predicting the unseen data.