In the Direction of an All-encompassing Comparison for Autism Diagnosis Using Assembles

Abstract

Abstract: Autism Spectrum Disorder (ASD) is a neurodevelopmental condition that affects social communication and behavioral patterns. Accurate and early diagnosis of ASD remains an important research challenge in medical data analysis and machine learning. While many previous studies have focused on individual (base) classifiers, the effectiveness of ensemble learning techniques for ASD diagnosis requires further investigation. In this study, a comprehensive comparative analysis of ensemble methods is conducted using the Adult Autism dataset obtained from the UCI Machine Learning Repository. The preprocessing stage includes handling missing values using the KNN imputation method and reducing the impact of outliers to improve data quality. Subsequently, four ensemble techniques, namely Bagging, Boosting, Voting, and Stacking, are applied and evaluated using WEKA and RapidMiner data mining tools. The performance of the proposed framework is assessed using Accuracy, Precision, Recall, and F1-score metrics. Experimental results demonstrate that ensemble learning methods provide highly competitive classification performance and outperform several corresponding base classifiers on the investigated dataset. The findings highlight the effectiveness of combining preprocessing techniques with ensemble learning approaches for improving ASD diagnosis and prediction.