Exploratory Data Analysis and Decision Tree Modeling for Autism Spectrum Disorder: Machine Learning Approach

Abstract

As the early diagnosis of autism spectrum disorder (ASD) is critical the high accuracy machine learning can be applied to achieve technology based diagnosis. In this context, the present study demonstrates machine learning approach for ASD diagnosis using decision tree (DT) modeling. The dataset employed in the present study comprises two classes of ASD adults with a sample size of 704 instances. The DT model entails a recursive partitioning approach implemented in the “rpart” package of R. The optimum model is derived by tuning parameters such as Min split, Min bucket, Max depth, and complexity. The performance of the model is evaluated in terms of the mean square error estimate of the error rate.