Autism Spectrum Disorder Diagnosis Using Machine Learning Algorithms

Abstract

The diagnosis of autism spectrum disorder (ASD) through conventional methods depends on behavioral assessments, which clinicians perform. The diagnostic process takes many clinical visits over extended periods, resulting in delayed early intervention opportunities. The therapeutic procedures that benefit patients create substantial emotional and logistical challenges for families. The research investigates how Machine Learning algorithms can improve the speed and accuracy of ASD diagnosis, creating a more efficient diagnostic process and enabling early, accessible interventions. The research evaluated three machine learning classification methods: K-nearest neighbors (KNN), stochastic gradient descent (SGD), and support vector machine (SVM). The models were trained and tested on an ASD-related feature dataset to assess their ability to identify individuals with ASD. The SGD classifier achieved the highest accuracy of 96% among the models. The high performance of ML-based diagnostic tools demonstrates their potential to improve traditional diagnostic methods by enhancing precision and efficiency. The research demonstrates that ML algorithms, specifically SGD, have strong capabilities for early and accurate ASD diagnosis. The implementation of these technologies reduces diagnostic delays and enables personalized intervention strategies, which lead to better outcomes for individuals with ASD while decreasing caregiver responsibilities.