Predicting Liver Cirrhosis Stages Using Extra Trees, Random Forest, and SVM with Data Mining Techniques

Abstract

Liver cirrhosis often occurs as a result of the lengthy and persistent progression of chronic liver disorders. It is a key crucial cause of death on a global scale. Early diagnosis and identification of cirrhosis are essential for preventing the disease's progression and the complete devastation of liver tissue. This paper aims to build an intelligent automated system that can predict the stages of cirrhosis employing Machine Learning (ML) algorithms, including Random Forest (RF), Extra Trees (ET), and Support Vector Machine (SVM). The dataset used in this research is sourced from the Zenodo website and linked to the GitHub website. This was our initial use of the data, which is publicly accessible and consists of 70 features and 10,000 records. In addition, data mining techniques were used to analyze the data before predicting the outcome. This involved data balancing due to the significant imbalance in the dataset's classes. To address this, we employed the Synthetic Minority Oversampling Technique (SMOTE) to mitigate a bias problem in a machine learning model. Then, feature selection techniques were applied, such as Chi-Square, Mutual Information (MI), and Recursive Feature Elimination and Cross-Validation (RFECV) based on classifiers RF and SVM (RF-RFECV, SVM-RFECV) to select relevant features. Lastly, the experimental findings showed that the Extra-Trees model with the Chi-square feature selection method (ET-Chi-Square) achieved the maximum level of accuracy of 93.87%. Additionally, it obtained recall, F1-score, and precision values of 94% each and an Area Under Curve (AUC) of 99%. Our method exhibited exceptional performance as compared to previous relevant research.