PAASD_HPT: Predictive Analytics for Autism Spectrum Disorder Prediction using Hyperparameter Tuning of Machine Learning Models

Background: Autism Spectrum Disorder (ASD) is a neurological condition which results in substantial communication, community, and behavioral challenges for those affected. Individuals with autism experience communication challenges, social interaction issues, and exhibit repetitive behaviors. Various machine learning techniques have been employed to distinguish persons with ASD from those without. Objectives: To design a hyperparameter tunning algorithm for machine learning model for predicting the ASD to enhance its overall efficiency. Methods: This study suggests Predictive Analytics for Autism Spectrum Disorder prediction using Hyper-Parameter Tuning (PAASD_HPT) of machine learning models. The Sine Cosine Algorithm (SCA) is employed to tune the hyper-parameter of different machine learning approaches including AdaBoost, Decision Tree, KNN, Random Forest and SVM. All classifiers' performance is evaluated using the ASD dataset, both prior to and following the hyperparameter adjusting process. Findings: The PAASD_HPT was tested on autism prediction in adults’ dataset from Kaggle. Empirical findings demonstrate that hyperparameter adjustment of machine learning models has diverse effects on predicting ASD. The proposed hyperparameter tuning approach achieves accuracy of 98.31% for AdaBoost, 96.62% for DT, 97.46% for KNN, 98.73% for RF and 96.20% for SVM. Upon hyperparameter optimization, Random Forest and AdaBoost surpasses alternative models with a notable statistical distinction. Novelty: The proposed PAASD_HPT utilized SCA to tune the hyperparameter. The novelty of this work is to optimize the machine learning models parameters using SCA. Thus, the suggested method helps to increase the accuracy while predicting the ASD using machine learning models. Keywords: Autism Spectrum Disorder (ASD), Classification, Predictive Analytics, Hyper-parameter tuning, ML Algorithms, Sine Cosine Algorithm