Efficient Mental Arithmetic Classification Using Approximate Entropy Features and Machine Learning Classifiers
Saif Al‐jumaili · DergiPark (Istanbul University) · 2023
In the current era, detecting mental workload is one of the most important methods used todetermine the mental state of humans, which in turn helps determine whether there is an issue inthe brain. Machine learning became the most used field used by researchers due to its accurateability to deal with and analyze the state of the brain. In this study, machine learning was used toclassify the Mental Arithmetic Task Performance (before and after) using EEG signals. Initially, as a preprocessing method, due to the variance of the signal received from the brain, we dividethe signal into Sub-bands namely alpha, beta, gamma, theta, and delta for artifact removal. Thenwe applied Approximate entropy (ApEn) to extract features from the signals. Next, the deducedfeatures were applied to 8 different types of classification methods, which are ensemble classifier, k-nearest neighbor (KNN), linear discriminate (LD), support vector machine (SVM), decisiontrees (DT), logistic regression (LR), neural network (NN), and quadratic discriminate (QD). Wehave achieved an optimal result using ES, furthermore, we compared our work with other papersin the literature, and the results outperformed them