|
|
 |
| |
|
|
|
Soft-Voting Ensemble Learning for Multimodal Rheumatoid Arthritis Classification |
|
|
|
PP: 1207-1218 |
|
|
doi:10.18576/amis/200506
|
|
|
|
Author(s) |
|
|
|
Sharik Ahmad,
Rafat Parveen,
|
|
|
|
Abstract |
|
|
| This article presents a multimodal machine learning (M3L) framework for the early prediction and classification of rheumatoid arthritis (RA) severity using a combination of clinical and biomedical data. Current approaches frequently rely on single models or unimodal datasets, which may reduce both accuracy and robustness. To overcome this challenge, the study introduces an ensemble learning method using a soft-voting classifier that integrates Gradient Boosting, Random Forest, and XGBoost. The proposed model determines RA severity stages based on informative features such as ESR, CRP, anti-CCP antibody levels, MRI joint inflammation indicators, gene expression scores, and additional relevant variables. By leveraging multiple algorithms, the framework mitigates overfitting and improves generalization capability. Performance is evaluated using accuracy, precision, recall, and F1-score, supported by comparative visual analyses to examine each model’s strengths and weaknesses. Results demonstrate that the ensemble achieves an accuracy of 0.88, outperforming XGBoost (0.86), Random Forest (0.86), and Gradient Boosting (0.87). |
|
|
|
|
|
|
|
