Predicting Happiness level using Machine Learning approaches among University Students in Bangladesh

Subjective wellbeing, or happiness, is a major and multi-faceted concept among university students that is linked to their mental health, academic achievement, and overall quality of life. Because of academic pressure, societal expectations, and the rapid environmental change in Bangladesh, it is very necessary for students to identify their psychological circumstances and level of happiness. The primary aim of this study was to develop a well-structured machine learning model for predicting happiness level among university students by detecting subtle trends in socio-demographic and psychological data to improve diagnostic accuracy and timely intervention. The sample was recruited from 500 university students using a multistage stratified cluster sampling approach. Additionally, the happiness level was measured using the Oxford Happiness Questionnaire (OHQ). Afterwards, the Boruta algorithm was implemented for selecting the most influential features. Moreover, various machine learning models were used to predict happiness levels. Finally, predictive performance was evaluated using various matrices, and nested cross-validation was implemented to ensure the model’s reliability and robustness. Among all the models, logistic regression exhibited the best predictive performance with the highest accuracy (63%) and balanced sensitivity (67.8%) and specificity (56.8%). Furthermore, the highest kappa score and a decent ROC value (59.99%) were found compared to all models. However, when maximizing sensitivity is the top priority, KNN can be a competitive alternative. The findings will serve as a strong foundation for educational institutions and legislators to create unique, research-based mental health support programs that will improve the impacted students' overall wellbeing and academic performance.