Optimization of Machine Learning Models for Early Diagnosis of Autism Spectrum Disorder

Abstract

Autism spectrum disorder is a syndrome related to interaction with people and repetitive behavior. ASD is diagnosed by health experts with the help of special practices that can be prolonged and costly. Researchers developed several ASD detection techniques by utilizing machine learning tools. ML provides the advanced algorithms that build automatic classification models. But disease prediction is a challenge for ML models due to the majority of the medical datasets including irrelevant features. Feature selection is a critical job in the predictive modeling for selecting a subset of significant features from the dataset. Recent feature selection techniques are using the optimization algorithms to improve the prediction rate of classification models. Most of the optimization algorithms make use of several controlling parameters that have to be tuned for improved productivity. In this chapter, a novel feature selection technique is proposed using binary teaching learning-based optimization algorithm that requires standard controlling parameters to acquire optimum features from ASD data.