Machine Learning Approach for Insomnia Early Diagnosis in Perspective of Bangladesh

Abstract

This research addresses the early detection of insomnia in Bangladesh, focusing on young adults affected by extensive phone and social media use. Using machine learning techniques like Logistic Regression, SVM, Decision Tree, Random Forest, XGBoost, CatBoost, Naive Bayes, and Light GBM, the study analyzed survey data from university students. The data processing was conducted using Python and Pandas, with null values handled carefully. Psychiatric validation was included. The models, especially Logistic Regression and CatBoost, achieved high accuracy (1.0 in Accuracy, ROC, AUC), suggesting a strong link between survey symptoms and insomnia. This approach, novel in Bangladesh, demonstrates machine learning's potential in mental health diagnosis, offering a cost-effective alternative to traditional methods. The study suggests further research to expand datasets and tailor models for diverse demographics, integrating these findings into public health policies.