Enhancing Deep Boosting Algorithm Performance for Panic Disorder Detection through Biometric and Spatiotemporal Data Integration

With the increasing proliferation of mobile devices, sensors, and Internet of Things (IoT) technologies, intensive research continues in the scientific community, driven by a commitment to promoting the well-being of the general population. It is the need of the hour to pay more attention to mental health as it directly impacts people’s lives. This paper aims to predict panic disorder ‘PD’ state by analyzing numerous health parameters in real time. To achieve this, the machine learning technique Bayesian-Light Gradient Boosting Machine hereafter referred to as ‘B-LGBM’ deep boosting is used to diagnose the health status of a person with PD. Hyperparameter optimization using the Bayesian technique is applied to identify the optimal set of parameters for the base learners, resulting in refined regularization values and reduced errors (overfitting) within the proposed model. Experiments on the open-source dataset reveal that the proposed B-LGBM model performed competitively with a mean square error ‘MSE’ score of 0.0364 and an accuracy of 96.36%. By combining artificial intelligence models with blockchain technology, future studies can enhance prediction accuracy and ensure secure, privacy-preserving assessment of complex physiological patterns in panic disorder. Our findings offer benefits in the areas of public mental health systems and clinical psychiatry in terms of tailored assessment and intervention.