Analisis Sentimen pada Teks Berbahasa Bali Menggunakan Metode Multinomial Naive Bayes dengan TF-IDF dan BoW

Abstract

Currently, digital technology is developing rapidly thus increasing the availability of textual data in the digital form. Many of the digital texts are available in Balinese. From the existing Balinese language texts, an analysis can be carried out to determine the emotional level or sentiment contained in them. Through this analysis, information will be obtained regarding sentiment towards a product or service so that it can be used as information for consideration in making decisions. To determine sentiment in textual data, more specifically unstructured data, several stages are required, one of which is feature extraction such as TF-IDF and BoW. This study will analyze the effect of TF-IDF and BoW feature extraction on the Multinomial Naive Bayes method. The test results show that the TF-IDF feature extraction provides precision, recall, accuracy, and F1-score values respectively 92.3%, 91.13%, 91.5%, and 91.38% higher than with BoW feature extraction.