Implementation of The Naïve Bayes Algorithm with Feature Selection using Genetic Algorithm for Sentiment Review Analysis of Fashion Online Companies

Abstract

Opinion rivalry that occurs in social media have an important role in increasing the potential customers to the company or agency. The review is a rich and useful resource for marketing, social and others for excavations and mining opinions such as views, moods, and behavior. The reviews describe perceptions of something, such as review of a product, review of airline services, reviews of restaurant and others. The analysis of sentiment is an ongoing field of text-based research. The analysis of sentiment or opinion mining is the study of ways to solve problems of public opinion, attitudes, and emotions of an entity, in which the entity may represent individuals, events or topics. Sentiment analysis is an important tool for analyzing opinions in social media. This measurement begins with pre-processing consisting of tokenizing, stopwords removal and stemming. This study uses naïve Bayes algorithm and genetic algorithms as applied feature selection. Selection features aim to classify text for the review of online fashion companies. This measurement results in the classification of text in form of positive text and negative text. Measurements are based on the accuracy of naïve Bayes before addition of genetic algorithms and after addition of genetic algorithms as feature selection. Validation using 10 fold cross-validation. For measurement accuracy using confusion matrix and ROC curve. The purpose of the study is to calculate the increased accuracy of naïve Bayes algorithm if using genetic algorithms for feature selection. The results showed that the genetic algorithm was able to improve the accuracy.