This research aims to analyze the opinion of the Twitter community regarding the election of Indonesian presidential candidates in 2024 using the Naive Bayes method. Data from 502 tweets was collected through scraping using Apify and then processed through the text preprocessing stage, including case folding, slang words, tokenizing, filtering, and stemming. The results of sentiment analysis using the Vader method showed 48 Negative sentiments, 30 Neutral sentiments and 424 Positive sentiments. Then, classification was carried out using the Naive Bayes model with an Accuracy of 77%. However, an imbalance was found in the label distribution which caused a bias towards the majority class. Negative Precission is 27%, Neutral Precission is 91%, and Positive Precission is 25%, while Negative Recall is 46%, Neutral Recall is 82%, and Positive Recall is 38%. Negative F1-Score reached 34%, Neutral F1-Score 87%, and Positive F1-Score 30%. These results show better performance for the Neutral class due to the larger amount of data, while the performance for the Negative and Positive classes tends to be low due to the lack of training data. This research provides insight into the public's mindset regarding registered presidential candidates and can serve as a guide for researchers, politicians and decision makers in understanding public views.

Penelitian ini bertujuan untuk menganalisis opini masyarakat Twitter terkait pemilihan calon presiden Indonesia tahun 2024 menggunakan metode Naive Bayes. Data dari 502 tweet dikumpulkan melalui scraping menggunakan Apify dan kemudian diproses melalui tahap preprocessing teks, termasuk case folding, slang word, tokenizing, filtering, dan stemming. Hasil analisis sentimen menggunakan metode Vader menunjukkan 48 sentimen negatif, 30 sentimen netral, dan 424 sentimen positif. Kemudian, dilakukan klasifikasi menggunakan model Naive Bayes dengan akurasi sebesar 77%. Namun, ditemukan ketidakseimbangan dalam distribusi label yang menyebabkan bias terhadap kelas mayoritas. Presisi negatif sebesar 27%, presisi netral 91%, dan presisi positif 25%, sedangkan Recall negatif sebesar 46%, Recall netral 82%, dan Recall positif 38%. F1-Score negatif mencapai 34%, F1-Score netral 87%, dan F1 Score positif 30%. Hasil ini menunjukkan kinerja yang lebih baik untuk kelas netral karena jumlah data yang lebih besar, sementara kinerja untuk kelas negatif dan positif cenderung rendah akibat kurangnya data pelatihan. Penelitian ini memberikan wawasan tentang pola pikir masyarakat terkait calon presiden yang terdaftar dan dapat menjadi pedoman bagi peneliti, politisi, dan pengambil keputusan dalam memahami pandangan publik.