Prediction of ailments using federated transfer learning and weight penalty-rational Tanh-RNN

The widespread adoption of technology has simplified the process of predicting ailments. Nevertheless, disease prediction for critical multi-organ systems was not done using the current approaches; rather, it was limited to single or dual organs. Therefore, utilizing WP-RT-RNN, a Federated Transfer Learning (FTL)-based disease prediction methodology is developed. First, the heart, lung, liver, and kidney dataset data are pre-processed, and then the Gaussian Mixture Model Hole Filling (GMMHF) technique is used for background subtraction. The image with the background removed is put into the Hurst operator, which extracts features. In the meantime, pre-processing is done on the EMR data before it is transmitted to the feature learning module based on Marginal Fisher Analysis-Convolution (MFA-CN), which effectively represents local and context features. Later that, the features gathered from the EMR data and images are combined and sent to the FSSTS-BCMO to choose the most important features. The Weight Penalty-Rational Tanh-Recurrent Neural Network (WP-RT-RNN) is then trained using the chosen features to effectively predict both normal and abnormal disease. At last, a performance comparison is carried out to confirm the suggested system’s efficacy.