Interpret Selected Features

After Boruta selection, use interpret_selected_features() to summarize confirmed features and cluster_correlated_features() to group redundant ones before final interpretation (Unterlerchner et al. 2023).

`interpret_selected_features()`

Builds a ranked table of confirmed features from a selection result dict (from select_relevant_features() or run_feature_extraction_and_selection_pipeline()).

Function Usage

python

interpret_selected_features(selection_result)

Entry Parameters

Parameter	Required	Type	Description
`selection_result`	✓	`dict`	Dict with `selected_feature_names`, `selected_indices`, and optionally `hit_counts` / `selected_mask`.

What It Returns

A DataFrame sorted by hit_count (descending, when available), then feature (ascending), with columns:

Column	Description
`feature`	Feature name.
`index`	Column index in the original `X` matrix.
`hit_count`	Boruta hit count vs shadows when the backend exposes it; usually `None` with BorutaPy.

With BorutaPy, hit_count is usually None—this is expected, not a bug. Use boruta_ranking, selected_*, and tentative_* from the selection result for diagnostics.

Example

python

from sequenzo import (
    run_feature_extraction_and_selection_pipeline,
    interpret_selected_features,
)

result = run_feature_extraction_and_selection_pipeline(
    seqdata=seqdata,
    outcome=outcome,
    controls=controls,
    preset="unterlerchner2023",
)

summary = interpret_selected_features(result)
print(summary)

`cluster_correlated_features()`

Groups selected features by hierarchical clustering on distance 1 - |corr|. Features with absolute Pearson correlation at or above abs_corr_threshold are merged into the same cluster.

Function Usage

python

cluster_correlated_features(
    X,
    feature_names,
    *,
    abs_corr_threshold=0.7,
)

Entry Parameters

Parameter	Required	Type	Description
`X`	✓	`ndarray` / `DataFrame`	Matrix of selected features only (`n` × `p_selected`).
`feature_names`	✓	sequence	Names matching columns of `X`.
`abs_corr_threshold`	✗	`float`	Merge features with \|corr\| ≥ threshold (default `0.7`).

What It Returns

A DataFrame with columns:

Column	Description
`feature`	Feature name.
`cluster_id`	Cluster assignment (0-based).
`mean_abs_corr_with_cluster`	Mean absolute correlation with other cluster members.
`representative_feature`	Feature chosen as cluster representative (highest mean link within cluster).

Sorted by cluster_id, then mean_abs_corr_with_cluster (descending), then feature.

Example

python

from sequenzo import (
    run_feature_extraction_and_selection_pipeline,
    cluster_correlated_features,
)

result = run_feature_extraction_and_selection_pipeline(
    seqdata=seqdata,
    outcome=outcome,
    controls=controls,
    preset="unterlerchner2023",
)

clusters = cluster_correlated_features(
    result["X_selected"],
    result["selected_feature_names"],
    abs_corr_threshold=0.7,
)

print(clusters[["feature", "cluster_id", "representative_feature"]])

R / Literature Mapping

Sequenzo	Literature
`cluster_correlated_features()`	Unterlerchner et al. (2023) correlation clustering before interpretation
`interpret_selected_features()`	Post-Boruta summary (no direct R function)

Notes

Pass selected features only to cluster_correlated_features()—typically result["X_selected"] and result["selected_feature_names"].
With BorutaPy, interpret_selected_features() may list hit_count as empty for every row; sort by boruta_ranking in the original selection result if you need a strength ordering.
Average linkage on 1 - |corr|; distance criterion t = 1 - abs_corr_threshold.
Empty input (p = 0) returns an empty DataFrame with the expected columns.
Single-feature input returns one row with cluster_id = 0.

Authors

Code: Yuqi Liang

Documentation: Yuqi Liang

References

Bolano, D., & Studer, M. (2020). The link between previous life trajectories and a later life outcome: A feature selection approach.

Unterlerchner, L., Studer, M., & Gomensoro, A. (2023). Back to the features. Investigating the relationship between educational pathways and income using sequence analysis and feature extraction and selection approach. Swiss journal of sociology, 49(2), 417-446.

Interpret Selected Features ​