Eye Movement Interpretation for Detecting Dyslexia Using Machine Learning Techniques

Abstract

Dyslexia is a learning disorder. It hampers an individual’s ability to comprehend the words, making it difficult for the person to read, spell, and write. The studies exemplify that timely detection and support proves as an advantage in mitigating the negative effects of dyslexia. the traditional manual method takes more time and effort to identify dyslexia, to overcome these issues the presented paper incorporated a machine learning based screening technique to detect dyslexia. As eye movements possess the ability to provide insights into reading disorders. Understanding the patterns that eye movements make while reading a text paragraph can help distinguish between dyslexic and non-dyslexic readers. The raw data consist of right-eye and left-eye movement positions along the x-axis and y-axis of 185 students were captured while reading a text paragraph. These eye movements were captured using an eye tracker based on the principle of human-computer interaction. The features such as fixation, saccadic movements were extracted for better prediction, later the classification was performed using XGBoost, support vector machine (SVM) and random forest (RF). The results show that XGBoost provides an accuracy of 95%, SVM 94% and RF 91%. To further validate the machine learning models author has used the performs measured called confusion matrix, precision, recall and F1 scores. The obtained results shows that the SVM achieved an F1 score of 94%, Recall of 94.5% and precision of 96%, whereas RF achieved an F1 score of 90%, Recall of 92% and precision of 89%. Finally, XGBoost achieved an F1 score of 95%, Recall of 95.5% and precision of 95%. The results imply that XGBoost achieves better result compare to other models.