Using Machine Learning Algorithms on Electroencephalographic Signals to Assess Engineering Students' Focus While Solving Math Exercises

Abstract

In this paper, we present an attention classification method using Machine-Learning Algorithms. The EEG signals were recorded from ten engineering students with an EPOC+BCI using the electrodes F3, F4, P7, and P8 while solving some mathematical operations. The recording time for these activities is around 20 minutes. Next, a similar time EEG register is obtained while doing non-academic activities, such as chattering with the staff, checking cell phones, or playing a video game. With these EEG registers, we obtained a set of features to train and evaluate attention using Machine Learning algorithms. This research shows how engineering students interact with math topics in solving mental operations and complex reasoning by increasing brain domain and knowledge for mathematical reasoning-related processes, such as sustained and shifting attention and logical constructions for object interaction during operations resolution. The Random Forest algorithm (RF) obtained the highest accuracy with 0.7392, an F1 Score of 0.7430, and the highest Specificity/Accuracy with 0.7261.