A novel modified long short term memory architecture for automatic liver disease prediction from patient records

The liver is the second largest organ in the human body after the skin and liver disease mainly impacts the liver's functionality by properly separating the nutrients and waste into the digestive system and also causes scarring (cirrhosis) as time passes. The scarring over time affects the healthy liver tissue and also affects its proper functioning and if left untreated for a prolonged period it can also result in severe complications such as liver failure or liver cancer. The patients can be prevented from the severe complications if the disease is detected at an earlier stage and the existing research for liver disease prediction mainly encouraged the usage of intelligent machine learning‐based techniques. However, these techniques have several complexities such as low accuracy, overfitting, higher training time, poor feature extraction capabilities and so on. To overcome these problems, we present modified long short term emory (MLSTM) architecture for chronic liver disease prediction. The proposed methodology has three stages: information enhancement, feature extraction, and classification. The modified generative adversarial network uses an autoencoder system for sample augmentation which helps to enrich the diversity present in both the normal and abnormal classes. The outlier information is eliminated via the criminal search algorithm which captures the differences and correlation associated with multiple samples. The fast independent component analysis algorithm and enhanced whale optimization algorithm are used for feature extraction. This step mainly identifies the crucial features for liver disease prediction and leaves out the irrelevant and duplicate features thus enhancing the convergence, computational time, and prediction accuracy. The MLSTM architecture is used to classify the samples present in the liver disease datasets into normal and abnormal (liver disease) classes. The proposed methodology offers improved performance in terms of accuracy, recall, means square error, and F‐measure. The results show that the proposed methodology will be efficient for doctors to diagnose liver disease in the earlier stage.