Enhanced Prediction of Autism Spectrum Disorder Using Kalman Filtering Based Support Vector Machine

Abstract

Autism Spectrum Disorder (ASD) is a category of complex neuro-developmental problems. ASD involves poor speech, limited desires, and repeated behavior. Early diagnosis of autism provides an opportunity an effective treatment. The challenges in correctly classifying and predicting ASD result in discouragement in seeking successful interventions. Machine learning techniques are being applied in medical fields for the prediction of diseases which would save precious time and cost. All machine learning algorithms won’t provide its better efficiency in predicting all diseases and it is limited. An algorithm that gives better performance in predicting a disease won’t give the same performance in predicting other diseases. This paper proposes an algorithm namely Kalman Filtering Based Support Vector Machine (KFSVM) to effectively predict the ASD. Optimum hyperplane in KFSVM assist better classification. To analyze the effective performance of KFSVM against previous algorithms it has been tested with three different ASD screening datasets available for adults, children and adolescents. Results are measured using benchmark data mining metrics and it has been found that KFSVM has better performance in classifying and predicting ASD in all considered datasets.