{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/visualizing-the-feature-importance-for-black","title":"Visualizing the Feature Importance for Black Box Models","arxiv_id":"1804.06620","date":"2018-04-18","proceeding":null,"authors":["Giuseppe Casalicchio","Christoph Molnar","Bernd Bischl"],"abstract":"In recent years, a large amount of model-agnostic methods to improve the\ntransparency, trustability and interpretability of machine learning models have\nbeen developed. We introduce local feature importance as a local version of a\nrecent model-agnostic global feature importance method. Based on local feature\nimportance, we propose two visual tools: partial importance (PI) and individual\nconditional importance (ICI) plots which visualize how changes in a feature\naffect the model performance on average, as well as for individual\nobservations. Our proposed methods are related to partial dependence (PD) and\nindividual conditional expectation (ICE) plots, but visualize the expected\n(conditional) feature importance instead of the expected (conditional)\nprediction. Furthermore, we show that averaging ICI curves across observations\nyields a PI curve, and integrating the PI curve with respect to the\ndistribution of the considered feature results in the global feature\nimportance. Another contribution of our paper is the Shapley feature\nimportance, which fairly distributes the overall performance of a model among\nthe features according to the marginal contributions and which can be used to\ncompare the feature importance across different models.","url_abs":"http://arxiv.org/abs/1804.06620v3","url_pdf":"http://arxiv.org/pdf/1804.06620v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"visualizing-the-feature-importance-for-black","repo_url":"https://github.com/giuseppec/featureImportance","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"feature-importance","task_name":"Feature Importance"}],"methods":[{"method_slug":"interpretability","method_name":"Interpretability"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1804.06620","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}