Benchmarking Variants of Recursive Feature Elimination: Insights from Predictive Tasks in Education and Healthcare
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Originally developed as an effective feature selection method in healthcare predictive analytics, Recursive Feature Elimination (RFE) has gained increasing popularity in Educational Data Mining (EDM) due to its ability to handle high-dimensional data and support interpretable modeling. Over time, various RFE variants have emerged, each introducing methodological enhancements. To help researchers better understand and apply RFE more effectively, this study organizes existing variants into four methodological categories: (1) integration with different machine learning models, (2) combinations of multiple feature importance metrics, (3) modifications to the original RFE process, and (4) hybridization with other feature selection or dimensionality reduction techniques. Rather than conducting a systematic review, we present a narrative synthesis supported by illustrative studies from EDM to demonstrate how different variants have been applied in practice. We also conduct an empirical evaluation of five representative RFE variants across two domains: a regression task using a large-scale educational dataset and a classification task using a clinical dataset on chronic heart failure. Our evaluation benchmarks predictive accuracy, feature selection stability, and runtime efficiency. Results show that the evaluation metrics vary significantly across RFE variants. For example, while RFE wrapped with tree-based models such as Random Forest and Extreme Gradient Boosting (XGBoost) yields strong predictive performance, these methods tend to retain large feature sets and incur high computational costs. In contrast, a variant known as Enhanced RFE achieves substantial feature reduction with only marginal accuracy loss, offering a favorable balance between efficiency and performance. These findings underscore the trade-offs among accuracy, interpretability, and computational cost across RFE variants, providing practical guidance for selecting the most appropriate algorithm based on domain-specific needs and constraints.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it