Optimization of Naïve Bayes Classifier Method Using Term Frequency-Inverse Document Frequency Approach (TF-IDF) Approach for Sentiment Analysis

Abstract

The main objective of this research is to conduct an analysis of public sentiment directed toward RSUD Siti Fatimah, using the Naïve Bayes Classifier methodology. This analytical approach was used to systematically categorize reviews into positive and negative sentiments. Data relating to the reviews was obtained through web scraping techniques from Google Maps, followed by a series of text preprocessing procedures, which included text sanitization, tokenization, and the application of TF-IDF for weighting. Based on the positive Classification values Precision shows 83%, Recal 1.00, and F-1 Score 0.91 which means the Model shows excellent performance in identifying positive sentiments. However, the model is less effective in identifying negative sentiments, with very low recall.