Autism Traits Detection in Toddlers: Enhanced Random Forest vs. Support Vector Machine

The aim of this study is to use behavioral indicators to identify autism in toddlers at an early stage. This will be done by applying a novel Random Forest (RF) Classifier and comparing its effectiveness with the Support Vector Machine (SVM) classifier. Both algorithms are evaluated using different training and testing splits to provide predictions about the features of Autism Spectrum Disorder in young children. The Gpower test, with parameters $\alpha=0.05$ and power $=0.80$, confirms an $80 \%$ confidence level. Results show that Novel RF achieves higher accuracy of $96.4020 \%$ compared to SVM of 90.5010% for the toddler dataset. An Independent sample t-test, conducted for statistical analysis, demonstrates a substantial disparity (p=0.007, p<0.05) in the performance of Novel RF and SVM. Thus, the accuracy of Novel RF outperforms SVM. Overall, Novel RF of $96.4020 \%$ demonstrates superior performance compared to SVM of 90.5010% in detecting autism traits in toddlers.