Traditional Machine Learning Fusion Model: An Efficient and Accurate Machine Learning Model for Antiviral Peptide Identification

The rapid mutation and spread of viral diseases have intensified the challenge of drug resistance to traditional antivirals, making the development of new antiviral agents crucial. Antiviral peptides (AVPs) have emerged as promising candidates due to their unique membrane penetration mechanisms and low resistance risk. However, conventional experimental screening methods are time-consuming and costly, while existing machine learning approaches suffer from limitations in feature representation and generalization capabilities. This study proposes an ensemble machine learning model, AVP, designed to identify antiviral peptides efficiently and accurately. The model integrates Random Forest (RF), Support Vector Machine (SVM), and Decision Tree (DT) classifiers using a soft-voting architecture with probability-based weighting. Regularization strategies, including L2 regularization for SVM and depth constraints for DT, are applied to enhance model stability. The model's performance is evaluated using five-fold cross-validation and ROC analysis. The AVP model achieves a training set AUC of 0.9980 and a test set AUC of 0.9784, demonstrating superior classification capability and generalization performance compared to traditional machine learning models. This study highlights the effectiveness of ensemble learning in fusing diverse feature-response patterns and provides a robust tool for antiviral peptide identification, accelerating the development of next-generation antiviral agents.