UNVEILING PUBLIC PERCEPTION OF TECHNICAL AND VOCATIONAL EDUCATION AND TRAINING (TVET) ON SOCIAL MEDIA USING A HYBRID SENTIMENT ANALYSIS APPROACH

Abstract

The public perception significantly impacts the effectiveness and societal acceptance of Technical and Vocational Education and Training (TVET) in Malaysia. Social media platforms such like Facebook becoming more important which makes it easier to get the real-time public input on education policy and programs. However, evaluating sentiment in informal and bilingual discussion poses significant challenges. This study presents a hybrid sentiment analysis approach that integrates lexicon-based approach (VADER, TextBlob, SentiWordNet, AFINN) with machine learning classifiers (Support Vector Machine, Naïve Bayes, Decision Tree, Logistic Regression, Random Forest and K-Nearest Neighbors) to classify and evaluate sentiments in 1,304 Facebook posts and comments related to TVET. The dataset was systematically pre-processed, translated, and subjected to feature extraction by Term Frequency-Inverse Document Frequency (TF-IDF). Accuracy served as the sole evaluation metric to ensure consistent comparison among lexicon-only, machine learning-only, and hybrid approaches. The hybrid combination of TextBlob and SVM achieved the highest accuracy at 72%, surpassing both standalone lexicon and machine learning approaches. This finding indicates that hybrid approaches are more effective in handling informal, code-switched sentiment data. The study offers a scalable framework for sentiment analysis in education policy, providing practical implications for TVET stakeholders aiming to evaluate public opinion and enhance engagement strategies.