Open AccessSentiment Analysis Using Naive Bayes Algorithm with Feature Selection Particle Swarm Optimization (PSO) and Genetic Algorithm
Author(s) -
Abi Rafdi,
Herman Mawengkang,
Syahril Efendi
Publication year - 2021
Publication title -
international journal of advances in data and information systems
Language(s) - English
Resource type - Journals
ISSN - 2721-3056
DOI - 10.25008/ijadis.v2i2.1224
Subject(s) - naive bayes classifier , feature selection , particle swarm optimization , confusion matrix , computer science , artificial intelligence , algorithm , feature (linguistics) , selection (genetic algorithm) , bayes' theorem , genetic algorithm , machine learning , pattern recognition (psychology) , bayesian probability , support vector machine , linguistics , philosophy
This study analyzes Sentiment to see opinions, points of view, judgments, attitudes, and emotions towards creatures and aspects expressed through texts. One of Social Media is like Twitter is one of the most widely used means of communication as a research topic. The main problem with sentiment analysis is voting and using the best feature options for maximum results. Either, the most widely known classification method is Naive Bayes. However, Naive Bayes is very sensitive to significant features. That way, in this test, a comparison of feature selection is carried out using Particle Swarm Optimization and Genetic Algorithm to improve the accuracy performance of the Naive Bayes algorithm. Analyses are performed by comparing before and after testing using feature selection. Validation uses a cross-validation technique, while the confusion matrix ??is appealed to measure accuracy. The results showed the highest increase for Naïve Bayes algorithm accuracy when using the feature selection of the Particle Swarm Optimization Algorithm from 60.26% to 77.50%, while the genetic algorithm from 60.26% to 70.71%. Therefore, the choice of the best characteristics is Particle Swarm Optimization which is superior with an increase in accuracy of 17.24%.