QUANTUM DRIVEN BAYESIAN OPTIMIZATION FOR LIVER CIRRHOSIS STAGE PREDICTION

Abstract

There are several ways to predict Liver cirrhosis stage (LCS) such as quantum computing (QC), machine learning (ML) and deep learning (DL), etc. This paper is dedicated to the use of Quantum Machine Learning (QML) to predict the conditions of LCS being either mild (ML), moderate (MD) or severe (SV). In particular, it will focus on using the Quantum Gradient Boosting (QGB) model, the Quantum Support Vector Machine (QSM) model and the Hybrid Quantum-Classical Model (HQM) to predict LCS.In addition, all the models are compared using metrics such as classification accuracy (A), precision (P), recall (R), F1 Score (F1S), training time (TT), memory usage (M) and model size (MS) by analyzing 80/20, 70/30 and 60/40 training/testing ratios prior and post the application of hyperparameter Bayesian optimization (HPO) tuning. To compare model performance metrics, the receiver operating characteristic (ROC) curve will be plotted based on true positive rate (TPR) and false positive rate (FPR) for each model and visualized using boxplots and heatmaps based on each metric. The findings suggest that HQM performs better than its competitors across all measures of A, P, R, F1S, and TT with average scores (95.53, 95.11, 95.25, 95.29 and 5.60) before BO, and (97.63, 97.52, 97.58, 97.55 and 5.46) after, therefore it  is clear that HQM outperforms the competition. However, QSM performs better in terms of  of M and MS with values 1275.19 and 89.31, and 1270.47 and 81.31 accordingly.