A1 Refereed original research article in a scientific journal
Estimating the prediction performance of spatial models via spatial k-fold cross validation
Authors: Pohjankukka J, Pahikkala T, Nevalainen P, Heikkonen J
Publisher: TAYLOR & FRANCIS LTD
Publication year: 2017
Journal: International Journal of Geographical Information Science
Journal name in source: INTERNATIONAL JOURNAL OF GEOGRAPHICAL INFORMATION SCIENCE
Journal acronym: INT J GEOGR INF SCI
Volume: 31
Issue: 10
First page : 2001
Last page: 2019
Number of pages: 19
ISSN: 1365-8816
DOI: https://doi.org/10.1080/13658816.2017.1346255
Self-archived copy’s web address: https://research.utu.fi/converis/portal/detail/Publication/26330595
In machine learning, one often assumes the data are independent when evaluating model performance. However, this rarely holds in practice. Geographic information datasets are an example where the data points have stronger dependencies among each other the closer they are geographically. This phenomenon known as spatial autocorrelation (SAC) causes the standard cross validation (CV) methods to produce optimistically biased prediction performance estimates for spatial models, which can result in increased costs and accidents in practical applications. To overcome this problem, we propose a modified version of the CV method called spatial k-fold cross validation (SKCV), which provides a useful estimate for model prediction performance without optimistic bias due to SAC. We test SKCV with three real-world cases involving open natural data showing that the estimates produced by the ordinary CV are up to 40% more optimistic than those of SKCV. Both regression and classification cases are considered in our experiments. In addition, we will show how the SKCV method can be applied as a criterion for selecting data sampling density for new research area.
Downloadable publication This is an electronic reprint of the original article. |  |
