Enhanced Early Diagnosis of Liver Diseases Using Feature Selection and Machine Learning Techniques on the Indian Liver Patient Dataset

Abstract

Liver diseases are a significant global health concern, with timely diagnosis crucial for effective treatment and prevention of further damage. This study addresses the challenge of early liver disease detection using machine learning techniques applied to the Indian Liver Patient Dataset (ILPD). Our proposed method comprises a four-phase approach: (1) initial model training using five machine learning algorithms - Multilayer Perceptron (MLP), Support Vector Machine (SVM), Decision Trees (DT-CART), Light Gradient Boosting Machine (LGBM), and Logistic Regression (LR) - on the original dataset; (2) feature selection using Forward Selection (FS) to identify the most relevant attributes; (3) model retraining with the selected features; and (4) model optimization to enhance prediction accuracy. The dataset was split into 80% training and 20% testing sets, with 10-fold cross-validation applied throughout. Our findings demonstrate the significant impact of feature selection and model optimization on algorithm performance. The Light Gradient Boosting Machine (LGBM) emerged as the top-performing model, achieving an accuracy of 82.12% after optimization, compared to its initial 76.21%. LGBM also showed balanced performance across specificity, sensitivity, precision, and F1-score metrics. This study contributes to the field by presenting a comprehensive approach to liver disease prediction, emphasizing the importance of feature selection and model optimization in improving diagnostic accuracy.