International Journal of Web Applications Ricardo Rodríguez Jorge 18 2 2026 https://doi.org/10.6025/ijwa/2026/18/2/39-65 https://www.dline.info/ijwa/fulltext/v18n2/ijwav18n2_1.pdf The rapid expansion of digital learning ecosystems and online educational platforms has generated substantial educational interaction data that can be utilized for intelligent forecasting and educational decision making. The COVID-19 pandemic further accelerated global dependence on online learning systems, thereby increasing the need for predictive analytical frameworks capable of understanding temporal educational behavior evolution. This study presents a machine learning based forecasting framework for predicting online learning activity trends using the dataset, “A.....Covid-19”. Dataset on Online Learning based Web Behavior from Different Countries Before and After COVID-19. The proposed framework integrates Random Forest Regression, Extreme Gradient Boosting (XGBoost), and Support Vector Regression (SVR) for both single-country and simultaneous multi country forecasting. Comparative benchmarking was conducted using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). The results demonstrate that XGBoost achieved superior forecasting performance because of its gradient-boosting optimization capability and nonlinear feature-learning efficiency. The developed framework further demonstrates the feasibility of AI-driven educational forecasting for large scale educational trend monitoring, digital infrastructure planning, and intelligent online learning analytics. The study also identifies several research directions involving explainable AI, deep learning forecasting architectures, and ethical educational intelligence systems.