An Emotional Intelligence Measures using Machine Learning Techniques for Higher Education

The demanding higher education system requires early detection of individuals at risk of low academic achievement or psychological discomfort. Academic metrics sometimes miss non-cognitive student success indicators. This study builds and tests machine learning models that predict student well-being using Emotional Intelligence (EI). A cross-sectional dataset of 800 undergraduate students was used to examine psychometric EI scores (Self-awareness, Self-regulation, Motivation, Empathy, Social skills), demographic characteristics, and sentiment analysis-based multimodal features. Four supervised learning models (Logistic Regression, Random Forest, Support Vector Machine, and Gradient Boosting) were examined for two classification problems: identifying students at Academic Risk (CGPA < 7.0) and predicting Help-Seeking Behaviour (counselling use). Gradient Boosting predicted Academic Risk better with a 0.91 F1 Score and 0.93 Recall on the test set. Most predictive was high academic stress, followed by Self-Regulation and Motivation. Multimodal sentiment analysis increased the model's recall to predict Help-Seeking Behaviour, exhibiting psychometric stability and emotional indicators