A Data-Driven Predictive Framework for Course Demand and Revenue Forecasting

This research presents a data-driven predictive framework for forecasting course enrollment demand and revenue on the EduPro online learning platform. The study develops an end-to-end analytical pipeline that transforms course, instructor, and transaction data into predictive insights for course planning, pricing, and revenue analysis. The framework uses a synthetic but schema-faithful dataset comprising 350 courses, 120 instructors, and 18,500 transaction records. It incorporates systematic data preparation and feature engineering, including price bands, course-duration buckets, rating tiers, instructor experience groups, and expertise-category matching. Five regression model families—Linear Regression, Ridge, Lasso, Random Forest, and Gradient Boosting—are benchmarked using an 80/20 train-test split and evaluated using MAE, RMSE, and R². Gradient Boosting demonstrates strong held-out performance, achieving an R² of 0.781 for enrollment prediction and 0.813 for revenue prediction. The study further examines feature importance to identify the major factors associated with enrollment demand and revenue. Course Rating is identified as the leading feature for enrollment prediction, while Course Price is the leading feature for revenue prediction, with Teacher Rating emerging as an important secondary factor. To support practical interpretation and decision-making, the project includes a structured KPI framework and an interactive Streamlit dashboard for presenting predictive results and analytical insights. The complete workflow demonstrates how feature engineering, model benchmarking, predictive analytics, and interactive visualization can be combined into a reproducible decision-support framework for online learning platforms. The work was developed as part of an internship project and provides a foundation for future enhancements including time-series forecasting, survival analysis, semantic expertise matching, uncertainty estimation, and automated performance alerts.