Hybrid model for precise hepatitis-C classification using improved random forest and SVM method.

Area of Science:

• Medical Informatics
• Machine Learning
• Bioinformatics

Background:

• Hepatitis C Virus (HCV) infection leads to liver inflammation, with millions of cases reported globally each year.
• Early diagnosis of HCV is critical for effective treatment and improved patient outcomes.
• Existing machine learning (ML) models for HCV prediction suffer from limitations like poor accuracy and data imbalance.

Purpose of the Study:

• To develop and evaluate a novel Hybrid Predictive Model (HPM) for Hepatitis C Virus (HCV) prediction.
• To overcome the limitations of existing single ML models in terms of accuracy and data imbalance.
• To enhance the performance of ML models in HCV diagnosis through feature selection and advanced techniques.

Main Methods:

• Proposed a Hybrid Predictive Model (HPM) integrating an improved Random Forest (IRF) with Support Vector Machine (SVM).
• Enhanced the Random Forest algorithm with a bootstrapping approach to iteratively eliminate minor features.
• Utilized a 'Ranker method' for feature selection and the Synthetic Minority Over-sampling Technique (SMOTE) to address dataset imbalance.

Main Results:

• The HPM achieved high accuracy rates, including 96.29% with 10-fold cross-validation and 92.39% with a 70:30 train-test split.
• Experiment 2 demonstrated a significant accuracy increase from 41.54% to 96.82% with SMOTE-based feature selection.
• The proposed HPM outperformed existing methods like SVM, MARS, RF, DT, and BGLM in accuracy.

Conclusions:

• The Hybrid Predictive Model (HPM) offers a robust and accurate solution for Hepatitis C Virus (HCV) prediction.
• Feature selection and techniques like SMOTE are vital for improving the performance of ML models in imbalanced datasets for HCV research.
• The study highlights the potential of advanced ML approaches for enhancing early HCV diagnosis and management.