Analysis of Cognitive Abilities in Students using Feature Optimization on EEG Signals

Abstract

The characterization of brain activity during cognitive load is a topic of growing interest in the scientific community. The electroencephalography technique has been extensively utilized for this purpose, providing valuable insights into the neural correlates of cognitive processes. In this work, EEG recordsobtained from the Physionet repositoryare analyzed, which are recorded while subjects performed mathematical tasks. The study divides the total 36 signals into two groups: "Good" and "Bad", potentially reflecting different levels of cognitive ability. Various temporal, frequency, and wavelet features were extracted from the EEG data using various signal processing techniques. These features were then classified using a range of machine learning techniques, including Multilayer Perceptron, Support Vector Machines,K-Nearest Neighbors, Linear Discriminant Analysis, and Naive-Bayes. Further the results compared with those obtained after applying feature optimization techniques, such as Particle Swarm Optimization,Genetic Algorithms, Firefly Algorithm, Sequential Floating Forward Selection, and Sequential Forward Selection. The experimental findings suggest that the KNN classifier optimized with FFA is particularly effective in characterizing brain activity under mental cognitive conditions with an accuracy of 95.17%, precision of 95.47%, recall of 91.52%, F1-score of 93.28%, and a False Positive rate of only 4.53%.The outcomes highlight the potential of proposed approach for understanding the neural mechanisms underlying cognitive abilities.