Ensemble of Customized Random Forest and Legacy Recurrent Neural Network with Fuzzy Learning Method Selector

Abstract

The rapid integration of Learning Management Systems (LMS) and Learning Analytics Dashboards (LAD) into digital education has transformed instructional delivery through scalable and data-driven platforms. However, these systems often face challenges such as rigid learning pathways, lack of personalization, and limited adaptability to diverse learner behaviors. To overcome these limitations, this work investigates and compares popular machine learning and neural network models in the LMS context. Experimental evaluations reveal that among traditional classifiers, Random Forest (RF) delivers superior performance, while Recurrent Neural Network (RNN) outperforms other neural models. Building on these findings, an intelligent ensemble framework is invented that comprises a Customized Random Forest (CRF) enhanced for contextual learning, a Legacy Recurrent Neural Network (LRNN) refined for temporal pattern recognition, and a Fuzzy Learning Method Selector (FLMS) that dynamically fuses both models based on fuzzy logic rules. The architecture is evaluated using key learning analytics parameters such as Accuracy, Precision, Sensitivity, Specificity, and F-Score to validate its predictive robustness and adaptability. The ensemble consistently outperforms baseline models including SVM, RF, ANN, and RNN, demonstrating its potential to elevate intelligent and personalized learning in LMS environments.