Research on Machine Learning-Based Prediction Models for Liver Cirrhosis Complicated by Hepatocellular Carcinoma
Year: 2025
Type: conference-paper
Study design: Observational Study
Abstract: This study aims to develop machine learning-based prediction models for hepatocellular carcinoma (HCC) development in liver cirrhosis patients, analyze potential influencing factors, and evaluate the impact of body weight through regression modeling, thereby providing evidence for clinical prevention and individualized treatment strategies. Utilizing clinical data from 419 cirrhosis patients at a specialized hospital in Fuzhou, we systematically compared the performance of Logistic Regression, Support Vector Machine (SVM), Artificial Neural Network (ANN), and Genetic Algorithm in model construction to optimize prediction accuracy and generalization capability. The results demonstrate that the SVM model achieved superior performance in HCC prediction, with an overall accuracy of 93.6%, precision of 90.1%, recall of 98.0%, and F1-score of 93.9%. The ANN model also exhibited strong predictive capability, showing training and test set AUC values of 0.86857 and 0.79826 respectively. In contrast, Logistic Regression underperformed in all evaluation metrics compared to SVM and ANN. Notably, the SVM model significantly enhanced screening sensitivity for high-risk HCC cases while demonstrating robustness against sample imbalance. This study successfully established and validated HCC prediction models based on SVM, Logistic Regression, and ANN, with particular emphasis on the superior predictive performance of the SVM framework. These comparative findings provide clinicians with enhanced tools for precise HCC risk assessment, facilitating early identification of high-risk patients and advancing the development of personalized therapeutic strategies.
Authors: Weice Wang, Xingchen Zhang, Hongyang Chen, Linqiang You, Yufeng Shen, Han Zhang, Yuxuan Lin, Yuting Huang, Changping Yang, Zhihui Huang
Institution: Yango University
Language: English
Subfield: Health Information Management
Field: Health Professions
Domain: Health Sciences
Open Access status: gold
OpenAlex ID: W4414931774