Optimizing ASD Diagnosis: Enhanced Autism Spectrum Disorder Prediction Using Machine Learning Classifiers

Autism Spectrum Disorder is a multidimensional neurological developmental disorder that includes a wide plethora of mental disorders targeting one's social and behavioral abilities. Gaining a perceptive understanding of a topic like ASD is essential in one's lifespan, whether the symptoms displayed are mild or severely debilitating. Due to the populous size of clinical terms under the disorder's spectrum and the drastic gap between each symptom's severity, the diagnosis of ASD became a long and tribulating task for experts and professionals alike. The purpose of this paper is to present and compare different Machine Learning Classifiers Categorical Boosting, K-Nearest Neighbors, Support Vector Machine, Logistic Regression, Decision Tree, Random Forest, Naïve Bayes and use their Training Algorithms Train-Test Splitting, K-Folds Cross-Validation to develop a quick and accurate Labeler ASD: “YES” ? “NO” for afflicted individuals based on multiple Datasets Generic, Adolescent, Children, Young, Adult, ASD aiming for an early diagnosis and an overall better understanding of the disorder. For each dataset, four Performance Metrics are extracted after applying the 7 ML models and the two training algorithms Accuracy, Precision, Recall, F1-Score. The resulting metrics are then displayed and compared to conclude the most accurate trained model for each dataset and with each of the training algorithms.