Utilizing grid search cross-validation with adaptive boosting for augmenting performance of machine learning models | JoVE Visualize

Area of Science:

Educational Technology
Computer Science
Data Science

Background:

The COVID-19 pandemic accelerated the adoption of Virtual Learning Environments (VLEs), generating vast amounts of student interaction data.
Effective analysis of this data requires appropriate Machine Learning (ML) algorithms for applications in Education Data Mining (EDM) and Learning Analytics (LA).
Selecting the optimal ML algorithm is crucial for accurately predicting student outcomes like performance, dropout rates, and engagement.

Purpose of the Study:

To improve the performance, accuracy, precision, recall, and F1 score of ML and Deep Learning (DL) algorithms in VLEs.
To identify and apply effective techniques for enhancing underperforming ML algorithms.
To determine feature importance for predicting student performance and inform the development of adaptive learning systems.

Main Methods:

Applied various ML algorithms to the Open University Learning Analytics (OULA) dataset.
Selected Decision Tree (DT) and Feed-Forward Neural Network (FFNN) algorithms for further experimentation due to unsatisfactory initial results.
Employed techniques including grid search cross-validation, adaptive boosting, extreme gradient boosting, early stopping, feature engineering, and neuron pruning to optimize algorithm performance.

Main Results:

Identified specific ML techniques that significantly improve performance metrics for predicting student study performance.
Determined the weights and importance of various features in the OULA dataset for predictive modeling.
Demonstrated the effectiveness of advanced ML optimization strategies on previously underperforming algorithms.

Conclusions:

Optimized ML algorithms can effectively process VLE data to predict student performance and engagement.
Feature engineering and ensemble methods are key to enhancing the predictive power of ML models in educational contexts.
The developed ML techniques and insights can guide instructors and administrators in optimizing learning content and providing adaptive support to students, thereby improving the overall learning experience.