A Semantic Approach of the Naïve Bayes Classification Algorithm

Abstract

This study aims to study the semantic approach of Naïve Bayes Classification Algorithm. From a statistical, probabilistic machine learning model, the classical decision-level classification algorithm which is the Naïve Bayes classifier shows to be efficient on a variety of sentiment classification problems. Naive Bayes is often used in sentiment classification applications and practical experiments because of its simplicity and effectiveness. However, its performance is often degraded because of the reliability of the result. This paper focuses on developing a different approach to the primary sentiment analysis of the NB classifier. The approach leads to the implementation of providing semantic information from lexicon resources together with the semantic calculator. Addressing the problems of the baseline algorithm show that it can be solved by incorporating other algorithm approaches. The comparative results show that the semantic approach is statistically superior and yielded improvements to the baseline classifier. In comparison with the baseline NB algorithm, SNB achieved a relatively favorable classification accuracy with the threshold of 50% to 60%and an average improvement of 11.394% for its accuracy rating while significantly reducing the training time.