#TITLE_ALTERNATIVE#

• Main Author: (NIM: 18213012), Hidayat
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/22448
• Institution: Institut Teknologi Bandung

The growth of machine learning technology in Indonesia, especially in text classification segment, goes rapidly. It comes with a lot of research that delivers many methods to optimize the performance of machine learning. To decide the method which has the best performance, one thing that should be done is to quantify the performance every single method. The quantifying process can be done by implementing them in sentiment analysis. Every feedback that’s been conveyed by Go-Jek’s customer in Twitter can be used as a dataset to analyze. Therefore, this research aims to decide which the alternative method best in doing sentiment analysis by using dataset about Go-Jek as tweet from Twitter in Indonesian. The approachment in text classification, generally, uses supervised learning. Some methods used in this research are Support Vector Machine, Naive Bayes Classifier, and Decision Tree Classifier. Every alternative methods used in this research is combined with weighting feature, such as TF and TF-IDF, and n-gram feature, such as unigram, bigram, and unigram+bigram. Naive Bayes Classifier combined with TF-IDF and unigram has best performance among the alternative methods used in this research with the 84% precision, 84% recall, and 84% F-Measure.