PENERAPAN ASPECT-BASED SENTIMENT ANALYSIS (ABSA) PADA KOMENTAR YOUTUBE UNTUK EVALUASI PERSEPSI PUBLIK TERHADAP MOBIL VINFAST MENGGUNAKAN METODE FINE-TUNING INDO-BERT

The automotive industry, particularly electric vehicles, is rapidly growing in Indonesia as public interest in environmentally friendly technology increases. In the digital era, online reviews serve as a crucial source for understanding public perception of a product. This study aims to analyze aspect-based sentiment from public opinions on VinFast electric vehicles marketed in Indonesia using the Fine-Tuning IndoBERT method. A total of 1,499 comments were collected from YouTube and, after filtering, 1,226 relevant comments remained. Six aspects were analyzed: performance, design, technology, price, battery, and comfort, which were manually labeled. The data underwent preprocessing and were split into 70% training data, 20% validation data, and 10% testing data. The model was trained using a multi-label classification architecture and evaluated for each aspect.
The results show that neutral sentiment dominates across all aspects (95.98%), followed by positive (1.97%) and negative (1.41%). The highest accuracy was achieved in the battery aspect (95%), while the lowest was in performance (66%). Positive sentiment was most prevalent in price, design, and performance aspects, whereas negative sentiment was highest in price and design, mainly related to complaints about high prices and battery costs. This approach proves effective in mapping public perception for each specific product aspect and can serve as a reference for manufacturers and policymakers in designing marketing strategies, product development, and policies to support electric vehicle adoption.
Keywords: Aspect-Based Sentiment Analysis, IndoBERT, Sentiment, VinFast, Electric Vehicle