Analysis of Enhanced Adaptive Gradient Boosting Regression in Business Intelligence

Abstract

Enhanced Adaptive Gradient Boosting (EAGB) is a real-time, streaming version of Gradient Boosting that updates weak learners’ one sample at a time and adapts automatically to concept drift through continuous error monitoring and dynamic model adjustments. It extends traditional gradient boosting by using incremental Hoeffding Tree regressors as base learners and updating the model continuously as new data arrive, rather than relying on static batch training. EAGB is particularly well-suited for Business Intelligence and forecasting tasks, including revenue prediction, demand forecasting, and financial analytics in streaming environments. Enhanced Adaptive Gradient Boosting Regression is important because it delivers consistently high predictive accuracy while effectively modeling complex and non-linear data patterns commonly found in real-world applications such as e-commerce, finance, healthcare, retail and others. The purpose of Business Intelligence is to attain desire’s optimum goal in business, rapid business decision making while minimizing time, resources consumption, and carbon footprint emission. The Enhanced Adaptive Gradient Boosting Regression in Business Intelligence helps organizations turn data into highly accurate predictions and insights, enabling better decisions and quicker responses. Enhanced Adaptive Gradient Boosting Regression in Business Intelligence serves vital role in addressing the real-world problems of businesses for example in retail to determine the sales of goods, revenue and services to consumers. Analyzing the Enhanced Adaptive Gradient Boosting Regression in Business Intelligence (BI) is crucial as it allows the organization to manage the trade-off between model performance, accuracy, cost, efficiency, resources consumption like CPU, RAM and emission of carbon dioxide from the computer’s system. The mechanism of Enhanced Adaptive Gradient Boosting Regression in Business Intelligence integrates sequential gradient boosting with online Hoeffding trees, residual clipping, EMA-based baseline updates, and ADWIN drift detection to enable continuous learning and incremental error correction. This design allows the model to adapt quickly to changing revenue patterns while maintaining stability during fluctuations, resilience to abrupt concept drifts, and efficiency in real-time processing of data. The novel aspect of this paper was that we developed a Business Intelligence model using Enhanced Adaptive Gradient Boosting Regression to solve the real-world business problems with accurate, quick response as well as better decision making in business. And also, we have analysed this model which compute performance metrics such as MSE, MAE, R2, and SMAPE, estimated system’s resources CPU, RAM usage, training time, energy consumption and estimated CO₂ emissions.