Experiments on malay short text classification

摘要

In this study, experiments are conducted on Malay short text using three diverse types of classifiers: KNN, SVM and NB. The classifiers were used to test the features of a bag-of-words (BOW) and a variant of TF.IDF; TF-IDF, smoothed TF-IDF and ITC. A Malay short text dataset was developed based on tweets from Twitter data and classified into two separate classes. The experiments were conducted on 50 % and 20 % sizes of the test data. The results demonstrated that the most highly consistent result was achieved by the SVM classifier with ITC as the feature, where the Precision, Recall, and F1-Score were all achieved at 95 %.