MD Mizanur Rahman

Work · Structured clinical data

Heart risk discrimination and calibration

  • Demo ready
  • Synthetic demonstration
  • Companion to coauthored research

Logistic regression, gradient boosting and a small MLP on a public structured dataset, with imputation and scaling fitted inside folds and calibration reported alongside AUROC.

Repository on GitHub

Problem

A model with a good AUROC can still be badly calibrated. On a 303-row dataset the difference between models is often inside the confidence interval.

Contribution

Relationship to research: Companion to coauthored research. New companion tabular benchmark for the coauthored heart-disease article; not that article's code.

Account owner’s role: Commissioned, reviewed and published the implementation. The code was produced with AI-assisted generation on commission and reviewed before release; it is not the original code of any cited publication.

What was built

  • Pipelines with in-fold imputation and scaling
  • Nested cross-validation
  • Brier score, calibration curves, threshold sensitivity, bootstrap intervals

Method

Threshold chosen on validation predictions only; no 2D reshaping of tabular inputs.

Data and access

UCI Heart Disease, Cleveland processed file (CC BY 4.0), obtained by the user; target mapped to num > 0.

Evaluation protocol

AUROC, AUPRC, Brier, expected calibration error with sample counts per fold.

Results

No measured result is published for this project. Demonstration outputs from the repository’s own synthetic fixtures are labelled as such in the repository and are not presented here as results.

Limitations

  • Educational benchmark; not a patient-facing risk service.

Reproducibility

Synthetic fixture with the same 13 columns supports the offline path.

Repository metadata: Python, MIT, last push 2026-09-13, 0 stars.

Source publication and links