Development of a Wilks feature importance method with improved variable rankings for supporting hydrological inference and modelling

Li, Kailong; Huang, Guohe; Baetz, Brian

doi:https://doi.org/10.5194/hess-25-4947-2021

Articles | Volume 25, issue 9

https://doi.org/10.5194/hess-25-4947-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/hess-25-4947-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 25, issue 9

Research article

|

09 Sep 2021

Research article |

| 09 Sep 2021

Development of a Wilks feature importance method with improved variable rankings for supporting hydrological inference and modelling

Kailong Li, Guohe Huang, and Brian Baetz

Supplement

https://doi.org/10.5194/hess-25-4947-2021-supplement

Data sets

Hourly data from surface meteorological stations in China China Meteorological Data Service Center http://data.cma.cn/en/?r=data/detail&dataCode=A.0012.0001

Data for "Gaining Hydrological Insights Through Wilk's Feature Importance: A Test-Statistic Interpretation method for Reliable and Robust Inference" Kailong Li https://doi.org/10.5281/zenodo.4387068

Short summary

We proposed a test statistic feature importance method to quantify the importance of predictor variables for random-forest-like models. The proposed method does not rely on any performance measures to evaluate variable rankings, which can thus result in unbiased variable rankings. The resulting variable rankings based on the proposed method could help random forest achieve its optimum predictive accuracy.