Electro-Search Algorithm and Autoencoder Based Recurrent Neural Network for Practical Medical Diagnosis

Nowadays, neural networks have a remarkable importance for medical applications. Because of their advantages such as dealing changing amount of data and resulting to higher accuracies at the end, they are widely used for especially medical diagnosis. At this point, there are different kinds of neural networks already provided successful results to the literature. In this study, it is explained that the classification rate (considering the medical diagnosis) can be increased by using intelligent optimization for feature selection and Autoencoder Based Recurrent Neural Network for performing the diagnosis. In this context, the Electro-Search Algorithm (ES) and Autoencoder Based Recurrent Neural Network (ARNN) have been employed for ensuring practical diagnosis over some known disease data sets from the active literature. The obtained findings by the ES-ARNN approach were compared with the findings by alternative techniques. Along the related evaluations, it was seen that the ES-ARNN resulted to high classification success rates for target medical diagnosis data sets and it also had higher diagnosing rates according to alternative techniques. The study also reports some real world experiences from physicians used the introduced solution over original data.