Insomnia Staging Through Facial Expression Analysis: A CNN-LSTM Hybrid Model

The goal of this research is to offer an in-depth examination of a classification model developed to categorize various health disorders using a combination of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. The efficacy of this model is assessed across several distinct categories in this study, with a particular emphasis placed on its capacity to identify and differentiate between a variety of health conditions reliably. The findings are laid out in the form of a comprehensive table, which demonstrates a variety of important performance measures for each class. These metrics include Certainty Rate, Recall, F1-Score, the number of scenarios, Class Occurrence Ratio, as well as Accuracy. The purpose of this research is to report on how well the model performed across four distinct health categories, namely “Healthy,” “Brief insomnia,” “Limited-term insomnia,” along “Persistent insomnia.” It can be deduced from the fact that the “Healthy” category achieved a Certainty Rate of 87.50%, a Recall Rate of 76.99%, with an F1-Score of 81.91 that the model correctly identified this class. Similarly, the three health classes, “Brief insomnia,” “Limited-term insomnia,” as well as “Persistent insomnia,” all displayed various levels of performance concerning these measures. In addition, the paper generates and offers many average performance indicators, such as Macro Average, Weighted Average, as well as Micro Average, which together provide a comprehensive picture of the overall performance of the model. The data presented in this study provide useful insights into the effectiveness of the CNN-LSTM model in diagnosing health disorders. The paper also highlights the model's strengths as well as areas in which it could be improved. This research is necessary for utilization in healthcare including medical diagnosis, both of which are fields in which a correct categorization of medical disorders is of the utmost significance.