Analisis Sentimen Polemik Kebijakan Pemerintah Makan Siang Gratis pada Twitter Menggunakan Metode Neural Network Classification

Abstract

This study aims to investigate public sentiment toward the Indonesian government's free lunch program by analyzing discussions on the X social media platform. A total of 1,500 tweets were collected through a web scraping process using relevant keywords and hashtags. The research workflow consisted of text preprocessing, including data cleaning, case folding, tokenization, stop-word removal, and stemming. The processed text was then transformed into numerical features using the Term Frequency–Inverse Document Frequency (TF-IDF) weighting method, followed by sentiment classification using the Neural Network Classification algorithm. Model performance was evaluated through K-Fold Cross Validation and a Confusion Matrix based on accuracy, precision, recall, and F1-score metrics. To address the issue of class imbalance, Random Over Sampling (ROS) was applied before the classification stage. The experimental results indicate that incorporating ROS improved the classification performance compared with the model trained on the original imbalanced dataset. Furthermore, the Neural Network Classification model effectively categorized public opinions into positive, negative, and neutral sentiments. The findings of this study are expected to provide valuable insights for policymakers in understanding public perceptions of the free lunch program and supporting future policy evaluation.