Regression-Driven Predictive Model to Estimate Learners' Performance through Multisource Data

Predictive modeling plays an essential role in comprehending and estimating the academic performance of learners within educational environments. This research introduces an innovative approach centered around regression analysis to forecast learners' achievements by harnessing data from various sources. By amalgamating a diverse array of data, including demographic particulars, previous academic records, lifestyle behaviors, forum participation, and engagement metrics, a holistic examination of factors influencing performance is achieved. The core of the predictive model is built on regression analysis, which capitalizes on the relationships between predictor variables and the target variable to establish a predictive framework. The methodology includes data collection, preprocessing, dataset partitioning, training using five regression algorithms (Random Forest, XGBoost, k-Nearest Neighbors, Gradient Boosting Regression Trees, and Support Vector Regression), and performance assessment using metrics like Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and, Accuracy. Results unveil insights into combined data-algorithm patterns, highlighting data source impact. Distinct categories' influence underscores data's multifaceted role in model accuracy. This underscores regression-driven models' potential for precise performance estimates, contributing to understanding the complex dynamics that drive successful online learning outcomes.