Identification of Autism Spectrum Disorder through Feature Selection-based Machine Learning

Abstract

Autism Spectrum Disorder (ASD) is a developmental disability that is likely to be perceived at a young age, persisting throughout a lifetime. The goal of this study is to detect ASD more efficiently with the use of Machine Learning methods. In our paper, we worked with the AQ-10 Adult dataset. Multiple steps have been taken to perform the data preprocessing. We have used different data synthesization techniques and a few feature selection techniques and eventually implemented them with other classifiers. Although throughout our analysis, we can see that the usage of Neural Network has some significant effect due to a smaller data set, the best-performance was provided by the combination of classifiers and feature selection methods to develop the prediction model. After evaluation, We deduced that a model with Principal Component Analysis (PCA) feature selection method using the AdaBoost classifier gave the best results.