9th International Conference on ICT for Sustainable Development

Authors - Cheritha Kondru, Abhinav Jayakuma, Sarath S
Abstract - This paper provides methods to recognize patients with potential Autism Spectrum Disorder(ASD). It is a tedious work to find it manually by going through every patient as it is troublesome. In this case, machine learning techniques are beneficial, helping people to diagnose ASD more quickly so that they don’t go through difficult procedures. Two datasets were chosen, one including adult-related information and the other toddler-related information. A number of models were trained on both datasets. The outcomes showed that the adult screening dataset performed better with the Support Vector Machine(SVM) (92.4%) and k-Nearest Neighbors(KNN) (97.1%) models when the accuracy was considered, and the toddler dataset performed well with the KNN (73.9%) and Gaussian Naive Bayes(GNB)(91.4%). With Receiver Operating Curve(ROC), KNN worked well with better results on the adult dataset and GNB for the toddler dataset.