Exploring Data Mining Techniques for Early Autism Detection using Random Forest Algorithm
M. Ayyadurai, K. Sujatha, R. Deeptha, D Preethi, M. Dinesh
Abstract
Autism, a neurological disorder, manifests uniquely in areas such as verbal and nonverbal communication, social interactions, behavioral adaptability, and specific interests. The results collected indicate that health professionals should receive more training in areas like psychology and developmental disorders and be able to detect Autism Spectrum Disorder (ASD) at a much earlier stage. They also provide pertinent information for reflection on the methods currently employed in evaluating children’s development and diagnosing ASD. This study collects information from a specific set of children with autism spectrum disorder from kaggle and is based on real, up-to-date facts. The input and output dataset will have certain common characteristics that will determine whether or not a child currently has ASD or has any signs of the same shortly. Random Forests, also known as Random Decision Forests, is a widely used ensemble learning method for tasks like classification and regression. This technique operates by constructing numerous decision trees during the training process. This classification is used to classify the dataset collected, into categories and hence segregate the children examined as “with” and “without” ASD with high accuracy about $88 \%$.