powershap
Feature selection using statistical significance of shap values
Decision gist · record as of 2026-08-14
Yes, with conditions. Powershap is a solid choice if you need statistically grounded feature selection and can tolerate aging maintenance (last update 323 days ago, no recent commits). The automatic mode removes hyperparameter tuning friction, and it integrates well with scikit-learn workflows. However, verify that its power-calculation defaults and statistical assumptions fit your problem domain before relying on it for critical feature selection decisions.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires Python 3.9 or later (supports up to 3.13).
- Runtime dependencies include catboost, pandas, scikit-learn, shap, and statsmodels.
- Low install friction with a pure Python wheel.
License · maintenance · safety
MIT (permissive) — MIT license permits commercial and private use with minimal restrictions, requiring only attribution and inclusion of the license text.
last release 2025-09-25 (323 days) · last repo commit 2025-10-07 · 216 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 117,711 downloads/mo, #12,151 on PyPI
Alternatives
Verify before relying
pip install powershap
from powershap import PowerShap
from catboost import CatBoostClassifier
X, y = ... # your classification dataset
selector = PowerShap(model=CatBoostClassifier(n_estimators=250, verbose=0, use_best_model=True))
selector.fit(X, y)
X_selected = selector.transform(X)- Whether automatic mode's power requirement default of 0.99 and false positive probability of 0.01 are appropriate for typical use cases.
- Performance and scalability characteristics on datasets with hundreds or thousands of features.
- How the method handles imbalanced classification or regression with heavy-tailed distributions.
What it is and what it does
Powershap is a feature selection method that combines Shapley value analysis with statistical hypothesis testing to identify which features are genuinely informative. It works by training multiple models on different data subsets, each time adding a random uniform feature as a baseline. For each feature, it calculates mean absolute Shapley values across iterations and compares them statistically to the random feature's impact using a percentile-based p-value test. Features with p-values below a threshold (default 0.01) are selected as significant.
The package includes an automatic mode that avoids manual hyperparameter tuning by using effect size and statistical power calculations to determine how many iterations are needed to achieve a target power level (default 0.99). It supports various model types—linear, tree-based, and deep learning—for both classification and regression, and integrates with scikit-learn conventions. The five runtime dependencies (catboost, pandas, scikit-learn, shap, statsmodels) are standard data science libraries.
Use it for
- Reduce dataset dimensionality before training a production model by identifying statistically significant predictive features.
- Compare feature importance across different model types to find consensus on which features matter most.
- Validate domain expertise by testing whether known important features rank above random noise in a statistical test.
- Automate feature selection in pipelines without manual threshold tuning, using the automatic mode's power-based iteration scheduling.
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes, with conditions.
Powershap is a solid choice if you need statistically grounded feature selection and can tolerate aging maintenance (last update 323 days ago, no recent commits). The automatic mode removes hyperparameter tuning friction, and it integrates well with scikit-learn workflows. However, verify that its power-calculation defaults and statistical assumptions fit your problem domain before relying on it for critical feature selection decisions.
Install
powershap on PyPI
Before you install
Low install friction with a pure Python wheel. Maintenance status is aging—last commit was 2025-10-07 and the package has not been updated in 323 days, though the repository remains active and not archived.
Requires Python 3.9 or later (supports up to 3.13). Runtime dependencies include catboost, pandas, scikit-learn, shap, and statsmodels.
License in practice
MIT license permits commercial and private use with minimal restrictions, requiring only attribution and inclusion of the license text.
Quickstart
pip install powershap
from powershap import PowerShap
from catboost import CatBoostClassifier
X, y = ... # your classification dataset
selector = PowerShap(model=CatBoostClassifier(n_estimators=250, verbose=0, use_best_model=True))
selector.fit(X, y)
X_selected = selector.transform(X)Verify before relying
- Whether automatic mode's power requirement default of 0.99 and false positive probability of 0.01 are appropriate for typical use cases.
- Performance and scalability characteristics on datasets with hundreds or thousands of features.
- How the method handles imbalanced classification or regression with heavy-tailed distributions.
Package facts
| License | MIT permissive |
| Python support | Supports the current Python release <=3.13,>=3.9 |
| Install friction | Low. Pure-Python wheel |
| Runtime dependencies | 5 packagescatboostpandasscikit-learnshapstatsmodels |
| Maintenance | Aging 323 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 117,711 / month, #12,151 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | Development Status :: 5 - Production/StableIntended Audience :: DevelopersLicense :: OSI Approved :: MIT LicenseOperating System :: MacOS :: MacOS XOperating System :: Microsoft :: WindowsOperating System :: POSIX :: LinuxProgramming Language :: PythonProgramming Language :: Python :: 3Programming Language :: Python :: 3.10Programming Language :: Python :: 3.11Programming Language :: Python :: 3.12Programming Language :: Python :: 3.13Programming Language :: Python :: 3.9 |
Evidence: powershap-0.1.0.1-py3-none-any.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “feature selection shapley values”
- powershapPowershap performs feature selection by testing whether each…
- shapSHAP computes Shapley values to explain individual predictions and…
- feature-engineFeature-engine provides transformers for engineering, selecting, and…
Give your agent the search over MCP, or paste the wish link into any chat.
More Artificial Intelligence packages
LiteLLM provides a unified Python interface to call 100+ LLM providers (OpenAI, Anthropic, Gemini, Bedrock, Azure, and others) using OpenAI-compatible API format, available as both a Python SDK and a self-hosted AI Gateway proxy server.
Install it if you need to work with multiple LLM providers or want to centralize LLM routing in your organization.
Client library and CLI tool for downloading, uploading, and managing models, datasets, and repositories on the Hugging Face Hub platform.
Install it if you work with Hugging Face Hub models or datasets.
LangChain provides a framework for building agents and LLM-powered applications by composing language models, tools, and memory through a unified API that abstracts over multiple model providers.
hf-xet provides chunk-based deduplication and efficient file transfer for the Hugging Face Hub, enabling faster uploads and downloads of large files with local disk caching.
Tokenizers converts raw text into token sequences for NLP models, with support for training custom vocabularies and using pre-built tokenizers (BPE, WordPiece) optimized for speed via Rust.
Transformers provides a unified framework for loading, fine-tuning, and running state-of-the-art pretrained models across text, vision, audio, video, and multimodal tasks using PyTorch, JAX, or TensorFlow.
Install it if you need to run or train any transformer-based model for NLP, vision, audio, or multimodal tasks.
See also Boruta · diptest · shap · powerlaw · tsfresh · hyppo · azureml-train-automl · aplr · bootstrapped · momentchi2