Abstract:
In this report, we aimed at using machine learning approaches to identify whether students’ self-esteem was low, normal, or high. Early identification of self-esteem was a strategic intervention in our model because of its key influence on mental health as well as performance outcomes. In questionnaires, we obtained basic demographic data, behavioral characteristics, and self-assessment results from 1,050 students. The results of the dataset showed that participants had low self-esteem in the range of 48.2%, normal self-esteem in the range of 50.9%, and only 1% of participants had high self esteem. In this paper, we were able to use machine learning algorithms, such as Logistic Regression, Random Forest, k-nearest Neighbors, and Gradient Boosting to build and train the developed model. To compare the effectiveness of each algorithm, we used parameters like precision, recall, F1-score, and accuracy in which Random Forest and Gradient Boosting models showed the best results for the classification of self-esteem categories. More details, hyperparameter tuning further enhanced the models to guarantee reliable predictive accuracy. That the use of machine learning has promising potential in mental health is supported by our studies which indicate how valuable self esteem assessments in an educational context may prove in terms of learning with early detection of the potential self-esteem issues at hand. Possible future research might provide a greater number of indicators, including social behavior and academic achievements to make the model more precise and suitable for study participants. From this research, it is possible that machine learning will play a significant role in creating practical evidence-based interventions to promote student well-being.
