ABSTRACT
Genetic disorders can be extremely dangerous if not detected and are difficult to identify. Patients can use genetic testing to help them make crucial decisions about the early diagnosis, treatment, or prevention of hereditary disorders. Genetic testing for these disorders must be done in infants and children since many children die from them. The current methods for identifying genetic abnormalities are very expensive, and some disorders are difficult to identify. Therefore, in this study, the authors suggest developing a system to identify the genetic condition that an individual possesses and also the person’s disorder type. In this work, multiple data preprocessing techniques are applied to the genetic disorder dataset taken from Kaggle. Autoencoders are used to extract a subset of input features because not all features are helpful for prediction. The genetic disorder is then classified using a variety of machine learning models, including OneVsRest Classifier, XGBoost, and artificial neural network (ANN). When the performances of these models are compared, ANN, with an accuracy rate of 94.13%, is regarded as the best choice. The performance of the resulting model is enhanced through hyperparameter tuning.
