An R package to analyze correlations and select a subset of non-redundant variables.
For details, see the post on Towards Data Science:
Fighting doppelgängers: How to rid data of evil twins reducing the feature space
dg <- doppelganger(data, priority={...}, threshold={...})
| Argument | Description |
|---|---|
| data | Data frame containing numerical variables. |
| variables | Columns of data to consider in the analysis (default: all). |
| priority | Ranking method to prioritize variables ("centrality", "peripherality", "raw_order"). |
| threshold | Correlation cut-off (absolute value). |
Variabiles to keep / drop:
dg$keep
dg$drop