Scholarlink

<p>Early prediction of autoimmune arthritis remains a critical medical challenge due to its complex pathophysiology and overlapping biomarkers. This study proposes a hybrid feature selection and classification framework, Cross Entropy in Objective Function for Metaheuristic Algorithms with XGBoost Classifier (CEOMAX), combined with an Iterative Mode-Based Consensus Feature Selection (IMCFS) mechanism to enhance predictive stability. CEOMAX integrates cross-entropy into the fitness function of metaheuristic optimization algorithms such as Particle Swarm Optimization (PSO), Grey Wolf Optimization (GWO), Genetic Algorithm (GA), Cuckoo Search (CS), and Jaya Algorithm (JA), which act as wrapper-based feature selection methods with the XGBoost classifier. IMCFS statistically refines the selected feature subsets by computing mode-consistent features across multiple stochastic runs, improving robustness. Experiments were conducted on the Arthritis Profile Data (APD), a clinically validated dataset containing 24 features and 102 patient samples collected from Sri Eswari Laboratory, India. In addition, the framework was evaluated on two benchmark datasets—Wisconsin Breast Cancer (WBC) and cardiovascular disease (CVD)—to demonstrate its generalization capability across different medical domains. The PSO-based CEOMAX variant (CEOPSOX) achieved an accuracy of 98.87%, and IMCFS further improved it to 99.25%. Overall, the CEOMAX–IMCFS framework improved classification accuracy by approximately 2–3% compared to misclassification-based optimization across the evaluated datasets. We evaluated the effectiveness of our proposed approach against other baseline and state-of-the-art methods for an overall performance comparison. Due to the small size and single-center nature of the APD dataset, the results should be viewed as exploratory and proof-of-concept. Validation on larger, multi-center datasets is required before drawing clinical conclusions. This work provides a computational pathway toward early autoimmune arthritis diagnosis and can be extended to larger, multi-center cohorts in future studies.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Prediction of autoimmune arthritis using cross entropy measures and iterative mode-based consensus feature selection method

  • 期刊论文

  • 发表日期：

    2026年6月3日

查看全文
  • Uma Ramasamy,
  • Sundar Santhoshkumar

摘要

Early prediction of autoimmune arthritis remains a critical medical challenge due to its complex pathophysiology and overlapping biomarkers. This study proposes a hybrid feature selection and classification framework, Cross Entropy in Objective Function for Metaheuristic Algorithms with XGBoost Classifier (CEOMAX), combined with an Iterative Mode-Based Consensus Feature Selection (IMCFS) mechanism to enhance predictive stability. CEOMAX integrates cross-entropy into the fitness function of metaheuristic optimization algorithms such as Particle Swarm Optimization (PSO), Grey Wolf Optimization (GWO), Genetic Algorithm (GA), Cuckoo Search (CS), and Jaya Algorithm (JA), which act as wrapper-based feature selection methods with the XGBoost classifier. IMCFS statistically refines the selected feature subsets by computing mode-consistent features across multiple stochastic runs, improving robustness. Experiments were conducted on the Arthritis Profile Data (APD), a clinically validated dataset containing 24 features and 102 patient samples collected from Sri Eswari Laboratory, India. In addition, the framework was evaluated on two benchmark datasets—Wisconsin Breast Cancer (WBC) and cardiovascular disease (CVD)—to demonstrate its generalization capability across different medical domains. The PSO-based CEOMAX variant (CEOPSOX) achieved an accuracy of 98.87%, and IMCFS further improved it to 99.25%. Overall, the CEOMAX–IMCFS framework improved classification accuracy by approximately 2–3% compared to misclassification-based optimization across the evaluated datasets. We evaluated the effectiveness of our proposed approach against other baseline and state-of-the-art methods for an overall performance comparison. Due to the small size and single-center nature of the APD dataset, the results should be viewed as exploratory and proof-of-concept. Validation on larger, multi-center datasets is required before drawing clinical conclusions. This work provides a computational pathway toward early autoimmune arthritis diagnosis and can be extended to larger, multi-center cohorts in future studies.