An exploratory analysis of longitudinal artificial intelligence for cognitive fatigue detection using neurophysiological based biosignal data.

Cognitive fatigue is a psychological condition characterized by opinions of fatigue and weakened cognitive functioning owing to constant stress. Cognitive fatigue is a critical condition that can significantly impair attention and performance, among other cognitive abilities. Monitoring this condition in real-world settings is crucial for detecting and managing adequate break periods. Bridging this research gap is significant, as it has substantial implications for developing more effectual and less intrusive wearable devices to track cognitive fatigue. Many models consider intricate biosignals, like electrooculogram (EOG), electroencephalogram (EEG), or detection of basic heart rate inconstancy parameters. Artificial Intelligence (AI)-driven methods aid in handling and categorizing these biosignals, recognizing fatigue-related patterns with higher accuracy. This technique is essential in high-demand surroundings such as education, healthcare, and workplaces or where cognitive fatigue may affect decision-making and performance. Therefore, the study presents an Exploratory Analysis of Longitudinal Artificial Intelligence for Cognitive Fatigue Detection Using Neurophysiological Based Biosignal Data (EALAI-CFDNBD) approach. The main aim of the EALAI-CFDNBD model is to detect cognitive fatigue using neurophysiological-based biosignal data. Primarily, the EALAI-CFDNBD model utilized the linear scaling normalization (LSN) model to ensure that the input features were appropriately scaled for subsequent analysis. Furthermore, the binary olympiad optimization algorithm (BOOA)-based feature selection is utilized to extract the most informative features, reducing the data dimensionality. The graph convolutional autoencoder (GCA) classifier is employed to classify cognitive fatigue detection. Finally, the multi-objective hippopotamus optimization (MOHO) method is utilized for parameter tuning, optimizing the model's hyperparameters to enhance overall detection accuracy. An extensive range of simulations is accomplished using the MEFAR dataset to establish a good classification outcome of the EALAI-CFDNBD method. The experimental validation of the EALAI-CFDNBD technique portrayed a superior accuracy value of 97.59% over the recent methods.