Prediction of Autism and Dyslexia Using Machine Learning and Clinical Data Balancing

Autism spectrum disorder (ASD) and dyslexia are expanding more swiftly than ever nowadays. Finding the characteristics of dyslexia and autism through screening tests is costly and time-consuming. Thanks to breakthroughs in artificial intelligence, computers, and machine learning, autism and dyslexia may be predicted at a very young age (ML). Even though several studies have been carried out using quite a few different approaches, none of them has shown a clear justification for how to predict autism and dyslexia traits across age groups. This study attempts to build a suitable prediction model enabled by ML technology to predict ASD and dyslexia for people of any age. This work seeks to examine the possible use of Random Forest, SVM with linear kernel, SVM with polynomial kernel, SVM with rbf kernel, SVM with sigmoid kernel, XGBoost, Decision Tree, Logistic Regression, Naïve Bayes, and KNN to forecast and assess ASD and dyslexia difficulties in children, adolescents and adults. Using real data set collected from individuals with and without autistic traits, the proposed model and the AQ-10 screening tool were assessed. The data for dyslexia is made up of 3644 cases with 197 properties, 196 of which are independent variables and one is a dependent variable. The data for autism consists of 704 cases with 22 characteristics, 21 independent variables, and 1 dependent variable with binary values (YES or NO). The results of the research showed that, in terms of accuracy, precision, F1 score, and recall, the recommended prediction model gave better results for the data set.