Extraction Sentiment Analysis Using naive Bayes Algorithm and Reducing Noise Word applied in Indonesian Language

Abstract

Sentiment Analysis is now very important and very useful in machine learning technology where a contextual mining of text to identify and extract subjective information in the source, and in helping to understand social sentiment from comments In general, sentiment analysis can be classified into three broad categories namely sentiment positive and negative. One method of machine learning is the Deep Belief Network (DBN). DBN which is included in the Deep Learning method, is by stacking several algorithms with several extraction features that utilize all resources optimally. This research has two points. First, it aims to classify positive, negative, and neutral sentiments for the test data. The following experiments provide a system of sentiment analysis through the naive Bayes algorithm to calculate sentiment and to improve accuracy by reducing noise in words applied in Indonesian language. From this research, a good level of accuracy can be obtained for extending sentiment using 10-Cross Validation resulting in an accuracy rate of 78.33% with an increase of 6.66% for systems that do not use stopwords, which means reducting noise words and the use of the naïve bayes classification method can be used to determine analysis sentiment on Indonesian language text.