Analisis Sentimen Film pada Twitter Berbahasa Indonesia Menggunakan Ensemble Features dan Naive Bayes

Rosy Indah Permatasari (Fakultas Ilmu Komputer, Universitas Brawijaya)
Mochammad Ali Fauzi (Fakultas Ilmu Komputer, Universitas Brawijaya)
Putra Pandu Adikara (Fakultas Ilmu Komputer, Universitas Brawijaya)
Eka Dewi Lukmana Sari (Sekolah Menengah Atas Negeri 6 Balikpapan)

Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer
Vol 2 No 11 (2018): November 2018

Sentiment analysis or opinion mining is one of the latest research topics in the field of information processing. It aims to know whether the polarity of a text-shaped data (document, sentence, paragraph) will lead to positive, negative, or neutral trait. This research used document text about Indonesian movie review which was obtained from Twitter. The method used in this research was Naive Bayes using Ensemble Features as a renewal feature beside Bag of Words Features. There are several types of Ensemble Features which are Twitter specific features, textual features, part of speech features, and lexicon based features. 500 data were used in this research, which were later divided into two types of data with the comparison of 70% for training data and 30% for testing data. The result of system accuracy obtained from sentiment analysis with Naive Bayes and Ensemble Features methods is 61.33%, 0.6369 precision, 0.5467 recall, and 0.5814 f-measure. The result of system accuracy using Ensemble Features and Bag of Words Features is 89.33%, 0.9041 precision, 0.88 recall, and 0.8922 f-measure.