Background: Autistic Spectrum Disorder (ASD) is a disorder associated with genetic and
neurological components leading to difficulties in social interaction and communication. According
to statistics of WHO, the number of patients diagnosed with ASD is gradually increasing. Most of
the current studies focus on clinical diagnosis, data collection and brain images analysis, but do not
focus on the diagnosis of ASD based on machine learning.
Objective: This study aims to classify ASD data to provide a quick, accessible and easy way to support
early diagnosis of ASD.
Methods: Three ASD datasets are used for children, adolescences and adults. To classify the ASD
data, we used the k-Nearest Neighbours method (kNN), the Support Vector Machine method (SVM)
and the Random Forests method (RF). In our experiments, the data was randomly split into training
and test sets. The parts of the data were randomly selected 100 times to test the classification methods.
Results: The final results were assessed by the average values. It is shown that SVM and RF are
effective methods for ASD classification. In particular, the RF method classified the data with an
accuracy of 100% for all above datasets.
Conclusion: The early diagnosis of ASD is critical. If the number of data samples is large enough,
we can achieve a high accuracy for machine learning-based ASD diagnosis. Among three classification
methods, RF achieves the best performance for ASD data classification.