Paper Title
Optimizing Weighted Voting for Real-Time Concept Drift Detection in Online Learning

Abstract
In online learning environments, concept drift is a serious problem since the underlying data distribution alters over time. Traditional learning models operate under the assumption of static data distributions, making them unsuitable to adapt to dynamic distributions, which can result in diminished predictive accuracy. This work presents an enhanced optimization technique for real-time concept drift detection in online learning, with a case study on airline passenger satisfaction. The work illustrates data distribution changes that can be accommodated by a hybrid ensemble of Random Forest and XGBoost. The proposed work implements Bayesian Optimization with the SLSQP approach to find the best weight distribution for each base classifier. The work continuously evaluates the accuracy and other performance metrics for the proposed approach & compares it with state- of- the art techniques. The proposed work achieve to an accuracy of 96.12%, precision of 96.15%, recall of 94.97%, and F1-score of 95.55%. Keywords - Concept Drift, Real-Time Online Learning, Adaptive Algorithms, Data Streams, Weighted Voting, Sequential Least Squares Quadratic Programming, Prudential Evaluation.