Expertise classification using functional brain networks and normalized transfer entropy of EEG in design applications

Expertise prediction is a challenging tasks for the development of state-of-the-art next generation computer-aided design (CAD) system. To develop an adaptive system that can accommodate the lack of expertise, the system needs to classify the expertise level. In this paper, we have presented a method to estimate the cognitive activity of the novice and expert user in the 3D modelling environment. The method has the capability of predicting novice and expert users. Normalized Transfer Entropy (NTE) of Electroencephalography (EEG) was used as a connectivity measure to calculate the information flow between the EEG electrodes. Functional brain networks (FBNs) were created from the NTE matrix and graph theory was used to analyze the complex network. The results from graph theory-based measures showed that there were significant differences between novice and expert user’s information flow patterns. The results showed that a classification accuracy of above 90% was achieved with a simple k-NN classifier and 5 features. From the feature selection method, we found that the most important EEG electrodes that contribute maximum towards classification were the frontal lobe electrodes. The classification results show that the proposed algorithm can effectively predict the novice and expert users in real-time.