AdaBoost meta-algorithm combining logit regression, neural networks, and bagged random forests¶

Element · All elements
Evidence · no claims cited

Description¶

The EDS's prediction engine combines three classifiers — a logit model, a multilayer perceptron neural network trained with backpropagation, and a bagged random forest using the C4.5 algorithm — via the AdaBoost boosting algorithm. The article states that boosting algorithms "evaluate the influence of the individual methods (weak classifiers) and merges the results into a single (strong) classifier," and that the AdaBoost prediction "has better prediction accuracy as compared to using any single method." This avoids depending on one method whose performance varies across universities and data availability.

Design Implications¶

Context¶

Requirements¶

• Individual base classifiers (logit, neural network, BRF) must first be trained on historical student data with known dropout/graduate outcomes

Constraints¶

• The article reports that within prior studies, different methods showed only marginal accuracy differences, and the most accurate method for a data set cannot be determined generally

Target Learners¶

• Bachelor students at German universities

Target Learning Goals¶

• Predicting dropout versus graduation to enable targeted intervention

Related Elements¶

• Early Detection System (EDS) built on standardized HStatG administrative student data

Examples¶

• Three-step attrition-combating process: identify at-risk students with administrative data, connect them to outreach programs, then evaluate the intervention

Key Sources¶

• Johannes Berens, Kerstin Schneider, Simon Görtz, Simon Oster, Julian Burghoff. (2019). Early Detection of Students at Risk - Predicting Student Dropouts Using Administrative Student Data from German Universities and Machine Learning Methods. Journal of Educational Data Mining, Volume 11, No 3. https://jedm.educationaldatamining.org/index.php/JEDM/article/view/389