PERFORMANCE ANALYSIS OF CLASSIFIER ON FACEBOOK DATA USING UNIGRAM & BIGRAM COMBINATIONS

Abstrak

This research on sentiment analysis uses the random forest method as classification. Tf-idf is a weighted feature and feature combination of  n-grams is unigram and bigram as feature words. In this research tf-idf used for the extraction feature, this test uses facebook comment data about the sports news.  In this study, datasets were used as much as 1000 data divided into 2, namely data testing and training data. Achieved  high accuracy performance results in unigram features with an accuracy of 83.67% of 2757 features,  bigram produces 58% with features as much as 8457.