Gallstone Disease Prediction with 38 Clinical Features and Model Performance

A tabular dataset used to study predictive models for gallstone disease presence. The dataset likely contains 38 clinical features, with model evaluation results showing a mean accuracy of 79.58% for a CatBoost classifier. The data was published by Prosenjit Das on figshare in June 2026.

Use Cases

• Benchmarking classification algorithms based on reported F1-score and accuracy metrics.
• Identifying key clinical predictors for gallstone disease based on the SHAP/LIME analysis mentioned.
• Studying feature selection techniques using the reported reduction from 38 to 19 features.
• Applying model explainability methods (SHAP, LIME, DiCE) to a clinical tabular dataset.

Strengths

• Includes specific model performance metrics (e.g., mean F1-score of 79.01%) for reproducibility.
• Model explainability results identify C-Reactive Protein (CRP) and Vitamin D as influential features.
• Released under a permissive CC-BY-4.0 license for open use.

Limitations

• Row count is unknown, which may limit suitability assessment for model training.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small (5.5 KB), indicating a limited scope or sample size.

Provenance

Source

figshare, author Prosenjit Das.

Collection Method

Method of original data collection is not described; the description focuses on model application.

Freshness

Last updated 2026-06-01 17:51:40; freshness should be verified.