Performance of Recurrent Neural Networks in Liver Disease Classification

Globally, it has been noted that more people are dying from liver disease than from other serious illnesses. The most prevalent forms of liver disease first exhibit no symptoms. This cause liver failure and lead to cancer. The dataset was acquired from a medical record for liver disease, which contains about 30,000 data and 10 attributes for identifying liver patients and non-patients. To get better accuracy for a huge dataset we implemented deep learning. The implementation was done with Recurrent Neural Network (RNN) algorithms LSTM, Bi-LSTM and GRU. By using all algorithms, we have got a test accuracy of 74.59% in LSTM, 85.41% in Bi-LSTM, and the highest accuracy 96.22% in the GRU model.