PARAMETERS OPTIMIZATION OF CONVOLUTIONAL NEURAL NETWORK USING PARTICLE SWARM OPTIMIZATION FOR MENTAL DISORDER DETECTION THROUGH FACIAL MICRO-EXPRESSIONS

Abstract

Schizophrenia is a prevalent mental illness with a significant global impact. Millions of individuals suffer due to the lack of early diagnosis and treatment. Detecting schizophrenia through facial micro-expressions offers a novel and promising approach. However, traditional deep learning methods often demand extensive computational resources and require complex hyperparameter tuning. This study proposes the use of Particle Swarm Optimization (PSO) to optimize the hyperparameters of Convolutional Neural Networks (CNNs), effectively replacing conventional backpropagation. PSO iteratively adjusts the parameters to find an optimal solution. Due to their computational efficiency and adaptability, CNNs are well-suited for deployment across various platforms. In this research, key CNN hyperparameters, including the number of epochs, batch size, number of filters, filter size, pooling size, and input pixel dimensions, are optimized. Additionally, PSO hyperparameters, including population size, social and cognitive weights, inertia weight, and the maximum number of iterations, are fine-tuned. The results show that the optimized CNN significantly outperforms conventional models in terms of reconstruction error and classification accuracy. This study presents an efficient method for optimizing CNN architectures, with applications in image compression, feature extraction, and anomaly detection. It also contributes to the early prediction of schizophrenia while maintaining low memory usage. The experimental results proved that the proposed CNN model gave the best accuracy values in comparison with other studies in the field. The proposed method can be considered as a powerful technique for prediction of schizophrenia mental disorder.