Comparative Analysis of Deep Learning-Based Hybrid Algorithms for Liver Disease Prediction

The liver is the largest and one of the most important organs inside the human body and its disorder affects the functioning of vital activities of the body. However, India alone has more than a million people diagnosed with liver diseases each year; hence, it is important to detect them at an early stage. We need an automatized process that can predict the symptoms of liver diseases and this can be done using machine learning algorithms or more specifically Deep Neural Network (DNN) algorithms. While these algorithms are used by researchers, their accuracies are low in the context of the security of humans being involved in the process. In this paper, we try to propose hybrid deep learning algorithms, which have higher degrees of accuracy. We have used three hybrid algorithms: CNN combined with LSTM (99.02%), CNN combined with GRU (98.38%), and CNN combined with RNN (99.48%). Experimental results establish that these accuracies, especially in the model where CNN is combined with RNN, are much higher than those of the existing algorithms in the literature. The methods used to achieve the results which involve preprocessing the data, feature selection, training, and testing followed by the use of the hybrid models, and then finally, the results obtained are analyzed using various metrics.