Evaluating The Impact of Social Media Sentiment on University Enrollment Decisions Using Machine Learning Classifier

Abstract

Public sentiment expressed through social media is increasingly recognized as a potential factor influencing higher education enrollment decisions. This study investigates whether sentiments on Twitter regarding Universitas Budi Luhur correlate with the number of new student admissions. To achieve this, tweet data were collected and analyzed using four supervised machine learning algorithms—Support Vector Classifier (SVC), Naïve Bayes, K-Nearest Neighbor (KNN), and Logistic Regression (LR)—combined with two lexicon-based sentiment dictionaries: SentiWord and InSet. Experimental results demonstrate that the SentiWord-based approach consistently outperformed the InSet-based approach across all models, with the SVC-SentiWord combination achieving the highest F1-score of 0.86. Despite the strong performance of these models in classifying sentiment, correlation analysis reveals no statistically significant relationship between Twitter sentiment and actual student enrollment trends. These findings underscore the effectiveness of lexicon-enhanced machine learning in sentiment analysis while raising important questions about the real-world impact of online sentiment on university admissions.