Perbandingan Metode Seleksi Fitur Pada Analisis Sentimen (Studi Kasus Opini PILKADA DKI 2017)

Abstract

In sentiment analysis, feature selection is a crucial step as it improves the performance and efficiency of sentiment analysis models. Feature selection also helps reduce the complexity of data dimensions, enabling faster and more efficient analysis. However, selecting relevant features poses a challenge as choosing the wrong features can decrease the accuracy of the constructed models. In this study, sentiment analysis was conducted on tweet data from the 2017 Jakarta gubernatorial election using TF-IDF feature selection combined with Recursive Feature Elimination (RFE), Chi Square, and Mutual Information. The models were evaluated using Naïve Bayes Classification (NBC) and Support Vector Machine (SVM) algorithms. Evaluation metrics such as accuracy, precision, recall, and F1-Score were used. The experimental results showed that the TfidfVectorizer + RFE combination in the NBC model achieved the highest accuracy of 71.1111% and demonstrated significant performance in terms of precision, recall, and F1-Score