Performance Analysis of Machine Learning Algorithms for Multi-class Document Using WEKA

Abstract

This research aims to assess and compare the performance of six machine-learning algorithms for text classification namely decision rules, decision tree, k-nearest neighbor (k-NN), naïve Bayes, regression, and Support Vector Machine (SVM). These six algorithms are compared using multi-class text document. The comparison was done in terms of their effectiveness, the ability of classifiers to classify the document in the right category. Precision, recall, F-measure, and accuracy are the four effectiveness measurements that were applied. The result shows that decision rule’s performance was the worst. SVM, decision tree, regression, and naïve Bayes have high effectiveness value. SVM can classify text quite well in average of 3.42 seconds to build each classifier model. Decision tree and regression can classify text with higher accuracy values rather than SVM, but slower in building the model. Among the six algorithms Naïve Bayes classifiers has the highest effectiveness value, while the model development time is the shortest as well. The average model building time is 0.03 second.