Predicting liver Disease Using Support Vector Machine and Logistic Regression classification Algorithm

Abstract

The liver plays a crucial role in various bodily functions, including protein production, blood clotting, and the metabolism of cholesterol, glucose, and iron. Early prediction of liver disease is vital for saving lives. In this study, machine learning algorithms were employed to predict liver disease, specifically Support Vector Machine (SVM) and Logistic Regression (LR), the primary aim is to provide an accurate, efficient, and non-invasive tool for early diagnosis and risk assessment of liver diseases. This allows for timely intervention, improved patient outcomes, and helps healthcare professionals make informed decisions. The dataset used was obtained from UCI (579 records). The models were trained with 405 samples (70%) and tested with 174 samples (30%). Key results showed that Logistic Regression outperformed SVM, achieving the highest accuracy of 97.24% and precision of 98%. SVM achieved an accuracy of 95.55% and a precision of 97%. Both models exhibited strong recall at 98% which indicates that both models are good for prediction of liver disease.  The models' convergence rates were 90 epochs for LR and 4750 epochs for SVM, indicating that LR converges much faster. These findings imply that Logistic Regression not only provides better predictive performance but also converges more efficiently, making it a more suitable algorithm for liver disease prediction