Chaos-Enhanced Grey Wolf Optimizer Integrated with LSTM

A Novel Chaos-Enhanced Grey Wolf Optimizer Integrated with LSTM for Improved EEG-Based Major Depressive Disorder Diagnosis

Major depressive disorder (MDD) is one of the leading mental health issues that has severe impacts on both individuals and society. Despite the potential of electroencephalographic (EEG) datasets to objectively diagnose MDD, the current limitation in feature extraction and optimization limits their use in practice. Traditional optimization algorithms often fail to find the right balance between exploration and exploitation, which makes them unsuitable for deriving significant patterns in electroencephalographic data. The paper develops a chaos-infused grey wolf optimizer with fractal enhancement (CIGWOFE), which is a hybrid framework that combines chaotic dynamics with the Grey Wolf Optimizer (GWO) and analyzes its performance alongside a long short-term memory (LSTM) network. The empirical findings show that CIGWOFE-LSTM significantly outperforms the state-of-the-art methods in terms of EEG-based major depressive disorder (MDD) classification, with a 17.4 % lower root mean square error (RMSE), a 12.6 % higher normalized squared error function (NSEF), and a 10.3% smaller bias. The model achieves a maximum classification accuracy of 94.5%, which is higher than GWO and equal to Runge-Kutta-GSS. The results of these performance improvements are significant, which is statistically confirmed by the Wilcoxon rank-sum test (p < 0.05). The proposed framework thus provides a sound and precise method for improving EEG-based diagnostics and reliably facilitates the detection of MDD in a clinical environment.