English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Shapley values originated in cooperative game theory but are extensively usedtoday as a model-agnostic explanation framework to explain predictions made bycomplex machine learning models in the industry and academia. There are severalalgorithmic approaches for computing different versions of Shapley valueexplanations. Here, we focus on conditional Shapley values for predictivemodels fitted to tabular data. Estimating precise conditional Shapley values isdifficult as they require the estimation of non-trivial conditionalexpectations. In this article, we develop new methods, extend earlier proposedapproaches, and systematize the new refined and existing methods into differentmethod classes for comparison and evaluation. The method classes use eitherMonte Carlo integration or regression to model the conditional expectations. Weconduct extensive simulation studies to evaluate how precisely the differentmethod classes estimate the conditional expectations, and thereby theconditional Shapley values, for different setups. We also apply the methods toseveral real-world data experiments and provide recommendations for when to usethe different method classes and approaches. Roughly speaking, we recommendusing parametric methods when we can specify the data distribution almostcorrectly, as they generally produce the most accurate Shapley valueexplanations. When the distribution is unknown, both generative methods andregression models with a similar form as the underlying predictive model aregood and stable options. Regression-based methods are often slow to train butproduce the Shapley value explanations quickly once trained. The vice versa istrue for Monte Carlo-based methods, making the different methods appropriate indifferent practical situations.

A Comparative Study of Methods for Estimating Conditional Shapley Values and When to Use Them