Correlation Filters in Machine Learning Algorithms to Choose Personal and Demographic Characteristics for the Diagnosis of Autism Spectrum Disorder
Keywords:
artificial intelligence, machine learning, autism spectrum disorder, data mining, diagnosis in neurodevelopmentAbstract
Currently regarded as one of the primary neurodevelopmental diseases, autism spectrum disorder is characterized
by restricted and repetitive patterns, as well as difficulties with social communication and cognitive abilities. There is no
known treatment for this illness, which varies in severity according on how each patient's symptoms manifest. The absence
of effective screening methods that are easy to use is one of the factors contributing to the often long wait times for autism
spectrum disorder diagnoses. The goal of this effort is to identify patterns of the autism spectrum by analyzing a public
database; in other words, to identify the traits that, when combined with behavioral traits, might increase the precursor model's
reliability. The probabilistic neural network approach did well in this categorization, according to the first findings.
Furthermore, the use of correlation filters showed increased accuracy and efficiency. We achieved 100% accuracy using the
support vector machine by using eight data mining techniques, combining the individual, behavioral, and demographic
factors, and eliminating certain features. Lastly, machine learning studies have shown that a patient's ethnicity, continent, and
presence of jaundice seem to indicate a higher likelihood of receiving an autistic spectrum disorder diagnosis.