Insomnia Detection using Machine Learning

Insomnia is a widespread sleep disorder that affects millions of people globally, posing significant risks to both mental and physical health. Early and accurate detection of insomnia is crucial to prevent its progression into more severe health concerns such as depression, cardiovascular diseases, and impaired cognitive function. However, traditional diagnostic approaches, relying heavily on clinical evaluations and subjective sleep diaries, are often time-consuming, costly, and susceptible to human errors. Recent improvements in machine learning (ML) have opened new ways for automating and enhancing insomnia detection, yet challenges such as model selection, and generalization across diverse populations persist. Although several ML models have been explored individually, systematic evaluation and comparison of classical and ensemble ML techniques for insomnia detection have not been emphasized. In this view, this paper proposes a comprehensive evaluation of a wide range of ML models, including Logistic Regression, Ridge Classifier, Random Forest, SVM, KNN, Naive Bayes, Gradient Boosting (GB), XGBoost (XGB), Neural Networks (NN), and ensemble combinations such as RF+NN, SVM+NN, GB+XGB, and KNN+NN. Extensive simulations were performed to validate the effectiveness of each model. The evaluation of performance was conducted across multiple metrics. Among all the models tested, Random Forest has shown superior performance.