Comparison of machine learning approaches to sentiment classification of Malaysian airline reviews
Muhammad Irham Abdul Razab, Haslizatul Fairuz Mohamed Hanum · UiTM Institutional Repositories (Universiti Teknologi MARA) · 2026
This study explores the use of machine learning (ML) methods to analyse customer reviews of Malaysia Airlines. The core problem is the need to correctly identify sentiment in unstructured online reviews, especially given language nuances, such as sarcasm, and the limited adaptability of prior models to Malaysia’s local, multilingual context. The main aim is to identify the most effective among four supervised ML models: Support Vector Machine (SVM), Logistic Regression, Naïve Bayes, and Random Forest (RF) to classify sentiment. Core aims include developing and training classifiers using TF-IDF and LDA-based feature extraction, and assessing performance using accuracy, recall, precision, and F1-score. The expectation is to find an optimal model for optimised sentiment analysis that can provide structured insights for airline operators. The study is limited to English-only text, excludes multimedia data, and uses moderately sized datasets.