A MACHINE LEARNING FRAMEWORK FOR EARLY STAGE DETECTION OF AUTISM SPECTRUM DISORDERS

Authors

  • Mr.VANIL KUMAR Author
  • SIDDHARTHA VALLAPOJU Author
  • REVOJU CHARAN Author
  • MINNIKANTI JAI SRIHARSHA VARDHAN Author
  • SIRIPURAM PRUDHVINATH Author

Keywords:

Autism Spectrum Disorder (ASD), Machine Learning (ML), Quantile Transformer (QT), Power Transformer (PT), Normalizer, Scaler (MAS)

Abstract

Autism Spectrum Disorder (ASD) presents significant challenges in affected
individuals' daily lives, and while complete eradication remains elusive, early
interventions can help mitigate its severity. This paper proposes a comprehensive
framework for evaluating various Machine Learning (ML) techniques for early ASD
detection. The framework employs four distinct Feature Scaling (FS) strategies:
Quantile Transformer (QT), Power Transformer (PT), Normalizer, and Max Abs
Scaler (MAS). Subsequently, the feature-scaled datasets undergo classification using
eight simple yet effective ML algorithms: Ada Boost (AB), Random Forest (RF),
Decision Tree (DT), K-Nearest Neighbors (KNN), Gaussian Naïve Bayes (GNB),
Logistic Regression (LR), Support Vector Machine (SVM), and Linear Discriminant
Analysis (LDA). Experiments are conducted on four standard ASD datasets covering
Toddlers, Adolescents, Children, and Adults. Evaluation measures including
Accuracy, Receiver Operating Characteristic (ROC) curve, F1-score, Precision,
Recall, Mathews Correlation Coefficient (MCC), Kappa score, and Log loss are
employed to compare classification outcomes and identify the best-performing
methods and FS techniques for each dataset. Results indicate AB achieving the
highest accuracy of 99.25% and 97.95% for Toddlers and Children, respectively,
while LDA achieves 97.12% and 99.03% for Adolescents and Adults datasets,
respectively. These accuracies are attained by scaling Toddlers and Children with
normalizer FS and Adolescents and Adults with the QT FS method. Additionally,
ASD risk factors are calculated, and key attributes are ranked using four Feature
Selection Techniques (FSTs): Info Gain Attribute Evaluator (IGAE), Gain Ratio
Attribute Evaluator (GRAE), Relief F Attribute Evaluator (RFAE), and Correlation
Attribute Evaluator (CAE). The study underscores the importance of fine-tuning MLmethods for predicting ASD across different age groups, providing valuable insights
for healthcare practitioners involved in ASD screening. Overall, the proposed
framework demonstrates promising results for early ASD detection compared to
existing approaches

Downloads

Published

08-06-2024

How to Cite

A MACHINE LEARNING FRAMEWORK FOR EARLY STAGE DETECTION OF AUTISM SPECTRUM DISORDERS. (2024). International Journal of Engineering Research and Science & Technology, 20(2), 552-558. https://ijerst.org/index.php/ijerst/article/view/320