Prediction of Autism spectrum disorder from high dimensional data using Machine Learning Techniques

Over the past few years due to the changes in the environmental and human lifestyle, several neurodevelopmental disorders are shooting their existence. Autism prevalence percentage is significant when compared to other neurodevelopmental disorders. To diagnose autism spectrum disorder medical experts, need big medical data and lot of time. Early detection of autism at an early age helps in curing the disorder. Machine Learning models are developed to predict Autism in high dimensional datasets. Dimensionality reduction is applied to identify features that effect Autism. Feature subsets are extracted using chi-square test, mutual information and lightgbm. Later the performances of the machine learning algorithms are analysed to find out suitable classifier that predicts the autism spectrum disorder. Using feature selection, the performance of Naïve Bayesian, K Nearest Neighbour has improved. Decision trees classifier performed the best with a performance score of 97.47%.