Classification of Indonesian Questions Using the Support Vector Machine Algorithm and Mutual Information Feature Selection

Abstract

Text classification can be used to organize, arrange and categorize a text. Text classification can be used for all text documents even if a text has a large number of features. However, the large number of features can cause reduced accuracy in the performance results of the classification system because there are some features that have less relevance to a text category. The Mutual Information feature selection method combined with the Support Vector Machine (SVM) algorithm is used to improve performance results in the classification process for Indonesian question documents by eliminating features with weights below the threshold. The results showed that the use of the Mutual Information feature selection method on the SVM classification algorithm was able to produce the best performance with an accuracy value of 0.92, precision: 0.93, recall: 0.89, f-measure: 0.9, computation time: 7 s and number of features: 240.   Keywords— Text Classification, Feature Selection, Support Vector Machine, Mutual Information