Improving the Accuracy of Multinomial Naïve-Bayes Algorithm with Adaptive Boosting Using Information Gain for Classification of Movie Reviews Sentiment Analysis

Abstract

Movie is a means of delivering information as well as entertainment that
can be enjoyed by all people through various platforms such as the
internet, cinema, and television. Sentiment analysis is needed to analyze
positive and negative comments from movie lovers, these comments
come from many circles and from various sources, one of which is
IMDb (Internet Movie Database). The naïve-Bayes multinomial
classification algorithm has been proposed and used by many
researchers in the case of sentiment analysis. The ensemble adaptive
boosting algorithm is used as a boosting algorithm to improve accuracy
in naïve-Bayes and information gain multinomial classification models.
The accuracy test on the model is carried out using the python
programming language. The accuracy results obtained when applying
the naïve-Bayes multinomial classification algorithm is 84.82%, then an
accuracy of 85.24% is obtained when implementing the information
gain feature selection on the naïve-Bayes multinomial classification
algorithm. The highest accuracy result of 87.87% was obtained when
implementing the naïve-Bayes multinomial classification algorithm with
adaptive boosting and information gain selection features.