Prediction of Liver Disease with Random Forest Classifier Through SMOTE-ENN Balancing

Abstract

Liver disease is a major public health issue that affects people all over the globe; its early diagnosis is an essential component of successful therapy and management. This study presents a comprehensive approach for liver disease detection-network (LDD-Net). The dataset consists of liver-related attributes, including demographic, clinical, and laboratory features. The preprocessing stage involves data cleaning, handling missing values, and normalization to ensure data quality and consistency. Additionally, skewness correction is applied to address the imbalance in the attribute distributions. Principal Component Analysis (PCA) is then employed for dimensionality reduction, extracting the most informative features while preserving the dataset's variance. The Synthetic Minority Over-Sampling Technique with Edited Nearest Neighbours (SMOTE-ENN) technique is employed for data balancing in order to correct the class imbalance that has been identified. SMOTE generates synthetic samples of the minority class, while ENN removes noisy and borderline instances. This combined approach aims to improve the classifier's performance by providing a balanced training dataset. Finally, the Random Forest classification (RFC) algorithm is employed to build a predictive model for liver disease detection. Random Forest leverages an ensemble of decision trees, offering robustness, scalability, and high accuracy in classification tasks. Experimental results on a liver disease dataset demonstrate the effectiveness of the proposed approach, which achieves promising results in terms of accuracy, precision, recall, and F1-score.

Topics & Concepts