Comparison of text classification methods

Abstract

The paper presents a comparison of some text categorization methods in terms of accuracy and learning speed. These methods are selected specifically for large dataset, therefore only the Random Forest algorithm is considered from the numerous machine learning techniques. In addition to this, two LSTM models are studied as – based on our literature review – these are found best suited to the text classification task among neural networks. Our research goal is to find evidence for or against this statement. Therefore we build, train and test a classic multilayer perceptron model and show its accuracy and learning speed as compared to the other methods.