Improving Feature Extraction for Sentiment Analysis on Indonesian Election 2024 Using Term Weighting with FastText

X or Twitter is widely used for communication and expressing opinions, often containing personal messages, opinions, and views on various topics, particularly around significant events like the Indonesian general elections. These public opinions can be leveraged to understand political trends, enhance services, de-velop marketing strategies, observe preferences, and even prevent adverse events. However, the unstructured and informal nature of tweets, poses a significant challenge for sentiment analysis and text classification. This research explores the improvement of text classification using the fastText framework coupled with term weighting techniques, focusing on the context of the Indonesian Election 2024 by reconfiguring the input layer of fastText, elim-inating irrelevant words from n-gram processed word vectors, and applying a term weighting technique that considers word occurrence and discriminating capacity. Experimental results demonstrate that the proposed enhancements lead to improved model performance, particularly with the TF-RF term weighting method. The TF-RF model achieved the highest accuracy (0.887), precision (0.918), recall (0.929), and F1-score (0.923), indicating its superior ability to make accurate and balanced predictions. These findings underscore the potential of optimized term weighting techniques to significantly improve fastText's performance in text classification tasks, providing valuable insights for future research and practical applications in social media analysis and political trend prediction.