Human Mental State Detection Using Modified Convolutional Neural Network with Leaky Rectified Linear Unit

In the present investigation, we explored the application of convolutional neural networks (CNNs) for classifying human mental states based on responses to questions related to mental health. The dataset includes responses aimed at evaluating various mental health conditions, providing a comprehensive basis for analysis. Our proposed CNNs architecture features multiple convolutional and pooling layers, followed by dense layers with leaky rectified linear unit (ReLU) activations to capture intricate patterns in the data. To prevent overfitting and ensure robust generalization, we integrate ridge regression (L2) regularization and dropout techniques during training. The model is optimized using the Adam optimizer and a binary cross-entropy loss function, striking a balance between computational efficiency and model performance. We evaluate the model's performance using standard metrics, including accuracy, confusion matrix, and classification report, achieving a notable accuracy of 90.24% on validation data. To gain deeper insights into the model's effectiveness, we employ receiver operating characteristic (ROC) curves and precision-recall curves, which illustrate the model's proficiency in identifying individuals with mental disorders while minimizing false positives and false negatives. Our findings highlight the significant potential of CNNs in mental health classification tasks, demonstrating substantial accuracy and generalization. This model represents a promising tool for the early detection and intervention of mental health disorders, potentially transforming healthcare outcomes and enhancing the quality of life for affected individuals. By providing a reliable method for mental health assessment, our research contributes to the advancement of mental healthcare, facilitating timely and effective support for those in need.