Text mining with probability-based classifier

Abstract

As the information which is controlled by the computer increases, it is made complicate to analyze and understand the information. Automatic Text Classification (TC) by their associates has greatly eased the control and processing of the massive volumes of information we face everyday. Among some of techniques used in handling the TC problems, we compare the Support Vector Machines (SVM) to the Relevance Vector Machines (RVM). The Support Vector Machine (SVM) is a decision machine so does not provide the posterior probabilities. While the Relevance Vector Machine (RVM) is relied on Bayesian formulation and provide the posterior probabilities. As the result of simulations, these two classifiers have similar performance measure except the number of selected vectors. The Relevance Vector Machine (RVM) is much sparser than the Support Vector Machine (SVM). So we can expect faster test time of the Relevance Vector Machine (RVM).