Intelligent Analysis Model of Behavior Decision Based on EEG Physiological Information

Abstract

It is increasingly crucial to build behavioral models with human physiological characteristics in complex scenes. Traditional methods are based on domain expert and computer-based knowledge discovery. This paper innovatively proposes a way of data acquisition that used pilot brain physiological signals for behavior modeling to solve the multiple flight sub-tasks classification. EEG pre-processing procedures based on data mining methods are proposed. Wavelet analysis and K-means algorithms are designed for feature extraction and optimization, respectively. The combination of quantitative and determinable modeling is adopted, which can help to understand the pilot behavior more accurately. Multiple sequence network coding scheme is used for quantitative physiological behavior model. Simultaneously, we proposed random forest algorithm to train a determinable behavioral decision model and grid search optimization should be required for hyper-parameter optimization. Results show that the alpha wavelet plays a significant role in behavioral decision-making. RF-GSO-Kmeans modeling obtain classification accuracy, precision, recall and f1-score indexes are 82.39%, 86.02%, 85.13%, and 85.22%. This will give important clues for future developments of behavioral model of robust and bionic characteristics combined physiological information and domain experts’ methods.