Machine learning model for predicting hepatocellular carcinoma in Hepatitis C patients

An estimated 58 million people suffer from chronic Hepatitis C virus around the world, while substantial evidence indicates that patients with Hepatitis C virus are at 17 times larger risk of developing Liver Cancer (Hepatocellular Carcinoma). Research has been carried out to predict hepatitis C and liver cancer at different stages in patients. In this research, we proposed the Classification model AdaBoost with Decision Tree as its base model to be trained and tested on patient dataset. The dataset contains records of clinical indicators and was acquired from the University of California Irvine Machine Learning Repository. The preparation of the dataset was done using balancing techniques i.e. SMOTE, it was encoded using Ordinal Encoding. The hyperparameters of AdaBoost model was tuned manually to find the most optimal combination. AdaBoost Classification Model achieved a 92.68% accuracy, and the AUROC of 0.97. The precision and recall differ for each class, “healthy” individuals were classified with a precision of 98% and a recall of 99% while patients with “Cirrhosis” (irreversible scarring of liver due to a tumor) were classified with 86% precision and 67% recall. The research concluded that machine learning has efficient applications in predicting diseases. Moreover, the clinical indicators mentioned in previous studies have proven to be vital in the prediction of HCC (liver cancer) however it is advised that, in future a larger dataset may be acquired to overcome any potential biases in the predictions. The current program successfully distinguishes between patients at different stages of HCC and HCV and can be further adapted to build decision systems to aid diagnosis.