The Impact of Feature Extraction to Naïve Bayes Based Sentiment Analysis on Review Dataset of Indihome Services

Indihome is a product of PT Telekomunikasi Indonesia as an internet service provider or internet service provider (ISP) in Indonesia. Every product or service offered to the public certainly has its advantages and disadvantages, as well as Indihome. From the advantages and disadvantages of Indihome services, we can do a technique, namely sentiment analysis. In this study, sentiment analysis was carried out regarding public responses or reviews about IndiHome services on Twitter social media. This study uses a comparison of TF-IDF and Word2Vec feature extraction, and the classification method used is the nave Bayes classifier. The accuracy results obtained in this study were 96% using the TF-IDF feature extraction and testing was carried out using an unseen data test that was selected randomly resulting in an accuracy of 92%. While the accuracy value obtained by using the Word2Vec feature extraction is 60% by testing using unseen test data that was selected randomly resulting in an accuracy value of 44%.