Meta-analytical Comparison Of SVM and KNN for Text Classification

Abstract

Text classification is a crucial method for Intelligent and AI based systems as the amount of text data increases from year to year. As sentiment analysis which is a prominent technique used by many companies and governments to understand the societies, it has become important to select the efficient and accurate text analysis algorithm to be used the text classification system. In this paper we present a meta-analytical study aimed at comparing two text classification machine learning algorithms namely KNN and SVM in terms of F-score. In addition to the meta-analytical study, our study presents a literature review for machine learning based text classification algorithms. For the meta-analysis, random and fixed models were used. The results of the meta-analysis using 95%-CI proved that there is no significant f1 performance difference between KNN and SVM in text classification tasks.