Dyslexia is a specific learning disorder where the individual often find difficulty in spelling and reading words fluently. Dyslexia is non-curable but with right remedial support, dyslexics can become highly successful in academics and life. Eye movement patterns during reading process can provide an in-depth understanding about reading disorders caused by dyslexia. Eye movements can be captured using eye-tracker, from which the relationship between how eyes move with respect to the words they read can be understood. In this work, a set of binocular fixation and saccade features were extracted from raw eye tracking data based on statistical measures. Machine learning algorithms such as Random Forest Classifier (RF), Support Vector Machine (SVM) for classification and K-Nearest Neighbor (KNN) were analyzed to output classification models for prediction of dyslexia. KNN gave higher levels of accuracy of 95% compared to SVM and RF over a small feature set of features related to fixations and saccades. These eye features can be used as a basis for developing screening means for prediction of dyslexia. Prediction of dyslexia at an early stage can help children to go for remediation which helps them for academic excellence.

Dyslexia is a learning disorder characterized by lack of reading and /or writing skills, difficulty in rapid word naming and also poor in spelling. Dyslexic individuals have great difficulty to read and interpret words or letters. Research work is carried out to classify dyslexic from non-dyslexics by various approaches such as machine learning, image processing, understanding the brain behavior through psychology, studying the differences in anatomy of brain. In addition to it several assistive tools are developed to support dyslexics. In this work, brain images are used for screening individuals who have high risk to dyslexia. This work also motivates the application of machine learning in distributed environment. The proposed predictive model uses the machine-learning algorithm Support Vector Machine (SVM). The model is designed in Apache SPARK framework to support voluminous data. The prediction accuracy of 92.5% is achieved using SVM.