Perbandingan Naive Bayes dan Support Vector Machine untuk Klasifikasi Sentimen Pengelolaan Sampah Banyumas

Abstract

Waste management is an important environmental issue because it is closely related to environmental cleanliness, public health, and quality of life. Public opinion on waste management is increasingly expressed through social media and produces large volumes of unstructured text data. This study aims to analyze public sentiment regarding waste management in Banyumas Regency and compare the performance of the Naive Bayes and Support Vector Machine algorithms in classifying sentiment into positive, negative, and neutral classes. The data were collected from X, YouTube, TikTok, and Threads using relevant keywords related to waste management in Banyumas. The dataset was processed through data selection, manual sentiment labeling, text preprocessing, feature extraction using TF-IDF, dataset splitting with an 80:20 ratio, and class imbalance handling using SMOTE. Model performance was evaluated using confusion matrix, accuracy, precision, recall, and F1-score. The results showed that Support Vector Machine achieved a higher accuracy of 68.63%, while Naive Bayes obtained 61.76%. However, Naive Bayes produced a better macro F1-score of 0.4573 compared to 0.3039 achieved by Support Vector Machine. These findings indicate that although Support Vector Machine performs better in recognizing the majority class, Naive Bayes provides more balanced performance across all sentiment classes. Therefore, Naive Bayes is considered more suitable for classifying imbalanced multi-platform sentiment data on waste management in Banyumas.

GS Cited Analysis