Enhanced Detection of Autism Spectrum Disorder using Machine Learning and Deep Neural Models with ADASYN and ENN Techniques

Autism Spectrum Disorder (ASD) is a neurodevelopmental condition that presents significant diagnostic challenges due to its heterogeneous characteristics and subtle behavioral patterns. In this paper, a hybrid approach is proposed to enhance the detection of ASD using ML and DL models, combined with intelligent techniques to address class imbalance and data noise. Three machine learning algorithms Logistic Regression, Decision Tree, and XGBoost are employed alongside two deep learning architectures: Artificial Neural Network (ANN) and 1D CNN. To improve the quality of the training data, Adaptive Synthetic Sampling (ADASYN) is used to generate synthetic minority samples, while Edited Nearest Neighbours (ENN) is applied to eliminate ambiguous or noisy instances. The models are trained and evaluated on original, ADASYN-adjusted, and ENN-filtered datasets using standard metrics such as accuracy, precision, recall, and F1-score. Results show that the combined use of ADASYN and ENN enhances model robustness, with deep learning models demonstrating superior performance in detecting ASD. This integrated framework holds promise for improving early diagnosis and facilitating timely interventions.