Accurate Prediction of Stage of Hepatitis C Virus Through a Stacking Ensemble

The current research involves the experimentation on HCV dataset through various machine learning pipelines so as to identify the most optimal one which yields the optimal values for the performance estimates. The HCV disease’s stage diagnosis aids in the appropriate treatment in a timely manner, thereby avoiding further damage to the patient’s health condition. Machine learning is employed on the dataset to perform multi-class classification over the labels indicating the stage of the disease wherein the pipeline includes the data preprocessing followed by the classification through various base classifiers and ensemble models. The dataset employed the dataset contains electronic health records of hepatitis C patients, contributed by Kanazawa University, Japan. Various linear and nonlinear classifiers were applied on the processed dataset including Naive Bayes’, Decision Trees, support vector machine, random forest, as base learners followed by their stacking ensemble. The predictive skill of the various machine learning pipelines is estimated through Repeated Stratified K-fold cross-validation technique. The performance metrics used are accuracy, precision, recall, F1-score, Mean Square Error, Mean Absolute Error, and Root Mean Square Error owing to the facts that dataset is heavily imbalanced, and it is a multi-class classification problem. It was observed that all the performance measures showed the optimal value with a stacking ensemble with F1-score having a value of 0.95 and Mean Square Error having a value of 0.06, respectively.