Comparative Analysis of Various Machine Learning Classifiers Applied for Genetic Disorder Prediction with Explainable AI

Genetic disorder is a disease caused entirely or partially by deviations from the normal DNA sequence. Hereditary gene mutations are linked to a number of well-known disease categories. When it comes to the prevention, treatment, or early identification of hereditary illnesses, genetic testing helps patients make critical decisions. Studies indicate that hereditary illnesses have increased exponentially with global population growth. According to WHO estimations, 10 out of 1000 people suffer from a genetic disorder [1]. Genetic illnesses affect the psychological, social, and physical well-being of patients and their families. Family dynamics are significantly impacted by genetic illnesses. They may need ongoing care and have no known cures or treatments, like many chronic illnesses. Hereditary illnesses are becoming more common due to limited awareness regarding the importance of genetic testing. This study addresses a significant gap in the prediction of genetic disorders quickly and accurately by presenting a robust ensemble framework that integrates four tree- and neural-based classifiers. In contrast to earlier studies that concentrate on a specific algorithm or do not provide a SHAP analysis, we make use of a publicly accessible cohort of 22,083 patient profiles - balanced using SMOTE.Our paper is focused on speeding up the diagnosis of genetic disorders so that the treatment needed can be administered earlier. To determine which classifier is more effective at predicting genetic disorders, We used various classifiers (such as Random Forest, SVM, XGBoost, AdaBoost, ANN, MLP, BERT, and Voting classifier). Our experiment's findings demonstrate the superiority of the Voting classifier, which has a higher prediction accuracy of 94.5%.