{"id":18521576,"url":"https://github.com/transferwise/shap-select","last_synced_at":"2025-07-30T07:34:09.574Z","repository":{"id":257808371,"uuid":"859750492","full_name":"transferwise/shap-select","owner":"transferwise","description":"A library for feature selection for gradient boosting models using regression on feature Shapley values","archived":false,"fork":false,"pushed_at":"2024-11-21T13:59:07.000Z","size":163,"stargazers_count":22,"open_issues_count":1,"forks_count":2,"subscribers_count":26,"default_branch":"main","last_synced_at":"2024-12-21T17:23:55.136Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/transferwise.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":".github/CODEOWNERS","security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-09-19T08:07:37.000Z","updated_at":"2024-12-12T04:57:04.000Z","dependencies_parsed_at":null,"dependency_job_id":"909601e7-23a0-41ff-8162-1a92a972e095","html_url":"https://github.com/transferwise/shap-select","commit_stats":null,"previous_names":["transferwise/shap-select"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/transferwise%2Fshap-select","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/transferwise%2Fshap-select/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/transferwise%2Fshap-select/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/transferwise%2Fshap-select/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/transferwise","download_url":"https://codeload.github.com/transferwise/shap-select/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":231294143,"owners_count":18354140,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2024-11-06T17:26:36.215Z","updated_at":"2024-12-25T23:42:12.996Z","avatar_url":"https://github.com/transferwise.png","language":"Python","readme":"## Overview\n`shap-select` implements a heuristic for fast feature selection, for tabular regression and classification models. \n\nThe basic idea is running a linear or logistic regression of the target on the Shapley values of \nthe original features, on the validation set,\ndiscarding the features with negative coefficients, and ranking/filtering the rest according to their \nstatistical significance. For motivation and details, refer to our [research paper](https://arxiv.org/abs/2410.06815) see the [example notebook](https://github.com/transferwise/shap-select/blob/main/docs/Quick%20feature%20selection%20through%20regression%20on%20Shapley%20values.ipynb)\n\nEarlier packages using Shapley values for feature selection exist, the advantages of this one are\n* Regression on the **validation set** to combat overfitting\n* Only a single fit of the original model needed\n* A single intuitive hyperparameter for feature selection: statistical significance\n* Bonferroni correction for multiclass classification\n* Address collinearity of (Shapley value) features by repeated (linear/logistic) regression\n\n## Usage\n```python\nfrom shap_select import shap_select\n# Here model is any model supported by the shap library, fitted on a different (train) dataset\n# Task can be regression, binary, or multiclass\nselected_features_df = shap_select(model, X_val, y_val, task=\"multiclass\", threshold=0.05)\n```\n\n\u003ctable id=\"T_694ab\"\u003e\n  \u003cthead\u003e\n    \u003ctr\u003e\n      \u003cth class=\"blank level0\" \u003e\u0026nbsp;\u003c/th\u003e\n      \u003cth id=\"T_694ab_level0_col0\" class=\"col_heading level0 col0\" \u003efeature name\u003c/th\u003e\n      \u003cth id=\"T_694ab_level0_col1\" class=\"col_heading level0 col1\" \u003et-value\u003c/th\u003e\n      \u003cth id=\"T_694ab_level0_col2\" class=\"col_heading level0 col2\" \u003estat.significance\u003c/th\u003e\n      \u003cth id=\"T_694ab_level0_col3\" class=\"col_heading level0 col3\" \u003ecoefficient\u003c/th\u003e\n      \u003cth id=\"T_694ab_level0_col4\" class=\"col_heading level0 col4\" \u003eselected\u003c/th\u003e\n    \u003c/tr\u003e\n  \u003c/thead\u003e\n  \u003ctbody\u003e\n    \u003ctr\u003e\n      \u003cth id=\"T_694ab_level0_row0\" class=\"row_heading level0 row0\" \u003e0\u003c/th\u003e\n      \u003ctd id=\"T_694ab_row0_col0\" class=\"data row0 col0\" \u003ex5\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row0_col1\" class=\"data row0 col1\" \u003e20.211299\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row0_col2\" class=\"data row0 col2\" \u003e0.000000\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row0_col3\" class=\"data row0 col3\" \u003e1.052030\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row0_col4\" class=\"data row0 col4\" \u003e1\u003c/td\u003e\n    \u003c/tr\u003e\n    \u003ctr\u003e\n      \u003cth id=\"T_694ab_level0_row1\" class=\"row_heading level0 row1\" \u003e1\u003c/th\u003e\n      \u003ctd id=\"T_694ab_row1_col0\" class=\"data row1 col0\" \u003ex4\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row1_col1\" class=\"data row1 col1\" \u003e18.315144\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row1_col2\" class=\"data row1 col2\" \u003e0.000000\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row1_col3\" class=\"data row1 col3\" \u003e0.952416\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row1_col4\" class=\"data row1 col4\" \u003e1\u003c/td\u003e\n    \u003c/tr\u003e\n    \u003ctr\u003e\n      \u003cth id=\"T_694ab_level0_row2\" class=\"row_heading level0 row2\" \u003e2\u003c/th\u003e\n      \u003ctd id=\"T_694ab_row2_col0\" class=\"data row2 col0\" \u003ex3\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row2_col1\" class=\"data row2 col1\" \u003e6.835690\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row2_col2\" class=\"data row2 col2\" \u003e0.000000\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row2_col3\" class=\"data row2 col3\" \u003e1.098154\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row2_col4\" class=\"data row2 col4\" \u003e1\u003c/td\u003e\n    \u003c/tr\u003e\n    \u003ctr\u003e\n      \u003cth id=\"T_694ab_level0_row3\" class=\"row_heading level0 row3\" \u003e3\u003c/th\u003e\n      \u003ctd id=\"T_694ab_row3_col0\" class=\"data row3 col0\" \u003ex2\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row3_col1\" class=\"data row3 col1\" \u003e6.457140\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row3_col2\" class=\"data row3 col2\" \u003e0.000000\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row3_col3\" class=\"data row3 col3\" \u003e1.044842\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row3_col4\" class=\"data row3 col4\" \u003e1\u003c/td\u003e\n    \u003c/tr\u003e\n    \u003ctr\u003e\n      \u003cth id=\"T_694ab_level0_row4\" class=\"row_heading level0 row4\" \u003e4\u003c/th\u003e\n      \u003ctd id=\"T_694ab_row4_col0\" class=\"data row4 col0\" \u003ex1\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row4_col1\" class=\"data row4 col1\" \u003e5.530556\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row4_col2\" class=\"data row4 col2\" \u003e0.000000\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row4_col3\" class=\"data row4 col3\" \u003e0.917242\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row4_col4\" class=\"data row4 col4\" \u003e1\u003c/td\u003e\n    \u003c/tr\u003e\n    \u003ctr\u003e\n      \u003cth id=\"T_694ab_level0_row5\" class=\"row_heading level0 row5\" \u003e5\u003c/th\u003e\n      \u003ctd id=\"T_694ab_row5_col0\" class=\"data row5 col0\" \u003ex6\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row5_col1\" class=\"data row5 col1\" \u003e2.390868\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row5_col2\" class=\"data row5 col2\" \u003e0.016827\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row5_col3\" class=\"data row5 col3\" \u003e1.497983\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row5_col4\" class=\"data row5 col4\" \u003e1\u003c/td\u003e\n    \u003c/tr\u003e\n    \u003ctr\u003e\n      \u003cth id=\"T_694ab_level0_row6\" class=\"row_heading level0 row6\" \u003e6\u003c/th\u003e\n      \u003ctd id=\"T_694ab_row6_col0\" class=\"data row6 col0\" \u003ex7\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row6_col1\" class=\"data row6 col1\" \u003e0.901098\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row6_col2\" class=\"data row6 col2\" \u003e0.367558\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row6_col3\" class=\"data row6 col3\" \u003e2.865508\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row6_col4\" class=\"data row6 col4\" \u003e0\u003c/td\u003e\n    \u003c/tr\u003e\n    \u003ctr\u003e\n      \u003cth id=\"T_694ab_level0_row7\" class=\"row_heading level0 row7\" \u003e7\u003c/th\u003e\n      \u003ctd id=\"T_694ab_row7_col0\" class=\"data row7 col0\" \u003ex8\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row7_col1\" class=\"data row7 col1\" \u003e0.563214\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row7_col2\" class=\"data row7 col2\" \u003e0.573302\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row7_col3\" class=\"data row7 col3\" \u003e1.933632\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row7_col4\" class=\"data row7 col4\" \u003e0\u003c/td\u003e\n    \u003c/tr\u003e\n    \u003ctr\u003e\n      \u003cth id=\"T_694ab_level0_row8\" class=\"row_heading level0 row8\" \u003e8\u003c/th\u003e\n      \u003ctd id=\"T_694ab_row8_col0\" class=\"data row8 col0\" \u003ex9\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row8_col1\" class=\"data row8 col1\" \u003e-1.607814\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row8_col2\" class=\"data row8 col2\" \u003e0.107908\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row8_col3\" class=\"data row8 col3\" \u003e-4.537098\u003c/td\u003e\n      \u003ctd id=\"T_694ab_row8_col4\" class=\"data row8 col4\" \u003e-1\u003c/td\u003e\n    \u003c/tr\u003e\n  \u003c/tbody\u003e\n\u003c/table\u003e\n\n\n## Citation\n\nIf you use `shap-select` in your research, please cite our paper:\n\n```bibtex\n@misc{kraev2024shapselectlightweightfeatureselection,\n      title={Shap-Select: Lightweight Feature Selection Using SHAP Values and Regression}, \n      author={Egor Kraev and Baran Koseoglu and Luca Traverso and Mohammed Topiwalla},\n      year={2024},\n      eprint={2410.06815},\n      archivePrefix={arXiv},\n      primaryClass={cs.LG},\n      url={https://arxiv.org/abs/2410.06815}, \n}","funding_links":[],"categories":[],"sub_categories":[],"project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftransferwise%2Fshap-select","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Ftransferwise%2Fshap-select","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftransferwise%2Fshap-select/lists"}