Optimal feature-algorithm combination research for EEG fatigue driving detection based on functional brain network

With the increasing number of motor vehicles globally, the casualties and property losses caused by traffic accidents are substantial worldwide. Traffic accidents caused by fatigue driving are also increasing year by year. In this article, the authors propose a functional brain network-based driving fatigue detection method and seek to combine features and algorithms with optimal effect. First, a simulated driving experiment is established to obtain EEG signal data from multiple subjects in a long-term monotonic cognitive task. Second, the correlation between each EEG signal channel is calculated using Pearson correlation coefficient to construct a functional brain network. Then, five functional brain network features (clustering coefficient, node degree, eccentricity, local efficiency, and characteristic path length) are extracted and combined to obtain a total of 26 features and eight machine learning algorithms (SVM, LR, DT, RF, KNN, LDA, ADB, GBM) are used as classifiers for fatigue detection respectively. Finally, the optimal combination of features and algorithms are obtained. The results show that the feature combination of node degree, local efficiency, and characteristic path length achieves the best classification accuracy of 92.92% in the logistic regression algorithm.