A predictive intelligence approach to classify brain–computer interface based eye state for smart living

Recently, brain–computer interface (BCI) based systems have become an emerging technology facilitating smart living. Accurate identification of eye states (open or closed) via an EEG-based BCI interface has many applications in a smart living environment, such as controlling devices and monitoring health status. Artificial neural networks (ANNs), including deep neural networks, are currently quite popular in many applications. In this study, a robust and unique ANN-based ensemble method is developed in which multiple ANNs are trained individually using different parts of the training data. The outcomes of each ANN are then combined using another ANN to enhance the predictive intelligence. The outcome of this ANN is considered the ultimate prediction of the user's eye state. The proposed ensemble method requires minimal training time and yields highly accurate eye state classification. An extensive analysis of bias and variance was used to assess the generalization ability of the proposed model while applying it to a real BCI environment and dataset. The proposed model outperforms traditional ANNs and other machine learning tools for eye state classification.