Breast Cancer Screening Prediction Eval

---
name: breast-cancer-screening-prediction-eval
description: Predicts county-level breast cancer screening rates using census-tract-level socioeconomic, demographic, and geospatial features. Evaluates the regression performance of Random Forest, Linear Regression, and Support Vector Machine models to identify which algorithm best captures underlying patterns in screening access disparities. Use when the user wants to benchmark on US Census Tracts Mammography Screening Dataset, or asks about evaluating this task. Reports R^2.
metadata:
  skill_kind: dataset_eval
  source_arxiv: 2502.06800
  bibtex_key: hashtarkhani2025geospatial
  confidence: high
---

# breast-cancer-screening-prediction-eval

> Analyzing Geospatial and Socioeconomic Disparities in Breast Cancer Screening Among Populations in the United States: Machine Learning Approach — Hashtarkhani et al. (2025) (arXiv:2502.06800, 2025)

## What this evaluates

Predicts county-level breast cancer screening rates using census-tract-level socioeconomic, demographic, and geospatial features. Evaluates the regression performance of Random Forest, Linear Regression, and Support Vector Machine models to identify which algorithm best captures underlying patterns in screening access disparities.

## Datasets

- **US Census Tracts Mammography Screening Dataset** — total 49118; splits: test (-1)

## Metrics

- `R^2` **(primary)** — range: [0, 1]
  - Coefficient of determination, measuring the proportion of variance in the mammography screening rate explained by the model. Calculated as 1 - (SS_res / SS_tot).
- `RMSE` — range: percent
  - Root mean squared error, measuring the square root of the average squared differences between predicted and actual screening rates.

## Input / output format

**Input**: Tabular features per census tract: location type (Rural/Urban), population density, demographic percentages (women ≥55, poverty, uninsured, higher education, Black, Hispanic), median home value, social vulnerability index, primary care shortage status, distance to nearest mammography facility, and number of facilities.

**Output**: Continuous predicted mammography screening rate (percentage).

## Scoring recipe

```python
def compute_metrics(y_true, y_pred):
    n = len(y_true)
    mean_true = sum(y_true) / n
    ss_res = sum((y - yp)**2 for y, yp in zip(y_true, y_pred))
    ss_tot = sum((y - mean_true)**2 for y in y_true)
    r2 = 1 - (ss_res / ss_tot)
    rmse = (ss_res / n) ** 0.5
    return r2, rmse
```

## Common pitfalls

- The paper displays county-level thematic maps but trains and evaluates models at the census-tract level, which may cause confusion about the unit of analysis.
- SHAP values are used exclusively for interpretability and feature importance/direction; they are not performance metrics and should not be confused with prediction accuracy.
- Several socioeconomic variables contain missing values (e.g., poverty rate, home value), but the exact imputation or exclusion strategy is not detailed in the provided text.

## Evidence (verbatim from paper)

> Evaluation of the final RF model, along with the LR and SVM models, based on $R^2$ and RMSE of the testing dataset is presented in Figure 2. The results indicate that the RF model, with an optimal number of trees set to 500 and the number of nodes (m) set to 4, outperforms both LR and SVM. Specifically, the RF model achieved a higher $R^2$ value and a lower RMSE, indicating its superior ability to capture and predict the underlying patterns in the data.

## Citation

```bibtex
@misc{hashtarkhani2025geospatial,
  title={Analyzing Geospatial and Socioeconomic Disparities in Breast Cancer Screening Among Populations in the United States: Machine Learning Approach},
  author={Hashtarkhani et al. (2025)},
  year={2025},
  note={arXiv:2502.06800}
}
```

- arXiv: 2502.06800