Neuro-Cognitive Mapping of Confusion States in Academic Environments

Abstract

In today’s digital education landscape, understanding student engagement and learning effectiveness 
has become increasingly vital. Traditional methods such as surveys, quizzes, or observational 
techniques fail to capture real-time cognitive states like confusion. This gap limits the ability to provide 
timely interventions, especially in asynchronous or large-scale e-learning environments. 
Electroencephalography (EEG) offers a promising solution by capturing real-time brain activity that 
can reflect cognitive states such as attention, meditation, and confusion. This Research aims to develop 
an intelligent system to detect student confusion using EEG signals and demographic data. The 
proposed system leverages machine learning algorithms, including Ridge Classifier, Linear 
Discriminant Analysis (LDA), and XGBoost, to classify student confusion during educational video 
interactions. It uses preprocessing techniques such as imputation, outlier handling, and feature scaling 
to ensure high-quality input data. Information Gain and correlation analysis are used for feature 
importance and selection. A Flask-based web application serves as the user interface, providing data 
visualization (EDA), model performance comparison, and real-time prediction functionality. Confusion 
metrics, feature importance plots, and EEG signal analysis are presented to enhance model transparency. 
The system also includes resampling strategies to handle class imbalance, ensuring reliable predictions. 
This Research demonstrates the feasibility of using physiological data and machine learning for 
adaptive education systems. It has significant implications for building intelligent tutoring systems that 
respond dynamically to individual learner needs, paving the way for more effective and inclusive digital 
learning environments.