ANALISIS SENTIMEN RESPON PENGGUNA CHAT GPT MENGGUNAKAN ALGORITMA SUPPORT VECTOR MACHINE

Abstract

This study aims to analyze user sentiment toward ChatGPT based on comments collected from the YouTube platform using the Support Vector Machine (SVM) algorithm. SVM belongs to the supervised learning algorithm group. The data were collected through web scraping using the YouTube Data API v3, resulting in 999 valid comments. The initial process included data cleaning using regular expressions to remove irrelevant characters, duplicates, and noise. Sentiment correction was then performed using a bilingual lexicon-based function (Indonesian and English) to improve classification accuracy based on language context. The initial sentiment distribution analysis showed 53.85% positive, 33.53% negative, and 12.61% neutral sentiments. To address class imbalance, a balancing process was conducted before model training. The preprocessing stage involved feature normalization and feature selection before splitting the dataset into 70% training and 30% testing data. The SVM model was trained and evaluated using performance metrics such as accuracy, precision, recall, F1-score, and AUC. The evaluation results showed an AUC of 0.90, accuracy of 81.6%, precision of 89.2%, recall of 51.6%, and F1-score of 65.4%. Based on these results, the SVM algorithm proved effective in classifying user sentiments toward ChatGPT with a high level of accuracy after the data balancing process.