Dyslexia is a reading disability and a language disorder where the individual exhibits difficulty in reading, writing, speaking, and trouble in spelling words. Early prediction of dyslexia can help dyslexics to get early support or intervention through remedial teaching. There is no remarkable computational model for the prediction of dyslexia in the literature. Existing methods to diagnose dyslexia include oral and written assessments, analysis and interpretation of Magnetic Resonance Imaging (MRI), functional MRI (fMRI), and Electroencephalogram (EEG). These methods require every instance to be interpreted by the domain expert in all stages whereas rigorously trained and tested computational models need subject expert intervention only at the end. In this paper, a prediction model has been proposed that uses statistical methods to differentiate dyslexics from non-dyslexics using their eye movement. The eye movements are tracked with an eye tracker. Eye movement has many features like fixations, saccades, transients, and distortions. From the raw data of eye tracker, high-level features are extracted using Principal Component Analysis. This paper proposes a Particle Swarm Optimization (PSO)-based Hybrid Kernel SVM-PSO for the prediction of dyslexia in individuals. The proposed model gives better predictive accuracy of 95% compared to a Linear SVM model. The proposed model is validated on 187 subjects by tracking their eye movements while reading. It is observed that eye movement data along with machine learning can be used for building models of high predictive accuracy. The proposed model can be used as a screening tool for the diagnosis of dyslexia in schools. © 2019, © 2019 IETE.