Identifies highly correlated feature pairs that may cause misleading
SHAP values or permutation importance.
Usage
xai_correlations(data, features = NULL, method = "spearman", threshold = 0.7)
Arguments
- data
Data frame of features
- features
Optional: subset of features to analyze
- method
Correlation method ("spearman", "pearson")
- threshold
Correlation threshold for warnings (default: 0.7)
Value
A list with correlation matrix, clusters, and warnings