Klasifikasi Sentimen Masyarakat terhadap Presiden Indonesia Menggunakan Metode Naive Bayes

Abstract. Social media platform X has become an important platform for expressing public opinion, particularly in the political context, including the 2024 Presidential Election in Indonesia. This study analyzes public sentiment toward the elected presidential pair, Prabowo Subianto and Gibran Rakabuming Raka, using the Naive Bayes algorithm. Research data was obtained through a scraping process on social media X, covering 1,478 tweets relevant to the research topic. The analysis process began with text preprocessing, which included cleansing, stemming, tokenizing, and stopword removal, followed by sentiment labeling into three categories: positive, neutral, and negative. The Naive Bayes algorithm was chosen for its simplicity, efficiency in processing large datasets, and ability to provide reliable results despite the assumption of feature independence. Additionally, the SMOTE method was applied to address data imbalance. The research results show that positive sentiment dominates with a percentage of 43.30%, followed by neutral sentiment at 36.47%, and negative sentiment at 20.23%. The application of SMOTE improved the algorithm's accuracy from 55.41% to 75.52%. These findings confirm that the Naive Bayes algorithm, with the support of data balancing techniques, is capable of providing reliable results in large-scale sentiment analysis. This research makes an important contribution to understanding public opinion on social media and serves as a guide in designing evidence-based policies. Abstrak. Media sosial X telah menjadi platform penting untuk mengekspresikan opini publik, khususnya dalam konteks politik, termasuk Pemilihan Presiden 2024 di Indonesia. Penelitian ini menganalisis sentimen masyarakat terhadap pasangan presiden terpilih, Prabowo Subianto dan Gibran Rakabuming Raka, dengan menggunakan algoritma Naive Bayes. Data penelitian diperoleh melalui proses scrapping pada media sosial X, mencakup 1.478 cuitan yang relevan dengan topik penelitian. Proses analisis dimulai dari text preprocessing yang meliputi cleansing, stemming, tokenizing, dan stopword removal, diikuti oleh pelabelan sentimen menjadi tiga kategori: positif, netral, dan negatif. Algoritma Naive Bayes dipilih karena keunggulannya dalam kesederhanaan, efisiensi pemrosesan data besar, dan kemampuannya memberikan hasil yang andal meskipun dengan asumsi independensi antarfitur. Selain itu, metode SMOTE diterapkan untuk mengatasi ketidakseimbangan data. Hasil penelitian menunjukkan bahwa sentimen positif mendominasi dengan persentase 43,30%, diikuti sentimen netral sebesar 36,47%, dan sentimen negatif sebesar 20,23%. Penerapan SMOTE meningkatkan akurasi algoritma dari 55,41% menjadi 75,52%. Temuan ini menegaskan bahwa algoritma Naive Bayes, dengan dukungan teknik penyeimbangan data, mampu memberikan hasil yang andal dalam analisis sentimen volume data besar. Penelitian ini memberikan kontribusi penting dalam memahami opini publik di media sosial serta menjadi panduan dalam merancang kebijakan berbasis data yang lebih responsif.