Optimized Framework for Mental Health Detection Using Deep Learning Technique

Psychological stress and mental health issues are growing concerns worldwide, requiring accurate and efficient detection methods. An optimized framework for mental health detection using a DL approach is proposed. The ECG5000 dataset was utilized for analysis, with the Weka Dl4jMlpClassifier employed to classify ECG signals. To categorize data into binary and multiple categories, the neural network structure has an input layer, two hidden layers, and an output layer. Important methods like ReLU activation, the Softmax function, cross-entropy loss, and the Adam optimiser were used to improve training accuracy and efficiency. With an accuracy of 90.8%, the new model exceeded the MDDBranchNet model’s accuracy of 76.0% on the same dataset. The enhanced accuracy of 14.8% shows the effectiveness of the proposed framework. The neural network showed good performance for the most common categories in the ECG5000 dataset. More effort needs to be done to improve the categorisation accuracy for less frequent categories. These results validate the ongoing development of effective deep learning-based approaches for the detection of mental health disorders.