4.7 Article

Spatial dependence between training and test sets: another pitfall of classification accuracy assessment in remote sensing

期刊

MACHINE LEARNING
卷 111, 期 7, 页码 2715-2740

出版社

SPRINGER
DOI: 10.1007/s10994-021-05972-1

关键词

Spatial autocorrelation; Cross-validation; Accuracy assessment; Overfitting; Remote sensing

资金

  1. French Ministry of Higher Education and Research (University of Toulouse)

向作者/读者索取更多资源

Spatial autocorrelation is inherent to remotely sensed data and can improve classification performance, but ignoring spatial dependence between training and test sets may lead to overestimation of generalisation capabilities. Spatial leave-one-out cross-validation is a better strategy for providing unbiased estimates of predictive error.
Spatial autocorrelation is inherent to remotely sensed data. Nearby pixels are more similar than distant ones. This property can help to improve the classification performance, by adding spatial or contextual features into the model. However, it can also lead to overestimation of generalisation capabilities, if the spatial dependence between training and test sets is ignored. In this paper, we review existing approaches that deal with spatial autocorrelation for image classification in remote sensing and demonstrate the importance of bias in accuracy metrics when spatial independence between the training and test sets is not respected. We compare three spatial and non-spatial cross-validation strategies at pixel and object levels and study how performances vary at different sample sizes. Experiments based on Sentinel-2 data for mapping two simple forest classes show that spatial leave-one-out cross-validation is the better strategy to provide unbiased estimates of predictive error. Its performance metrics are consistent with the real quality of the resulting map contrary to traditional non-spatial cross-validation that overestimates accuracy. This highlight the need to change practices in classification accuracy assessment. To encourage it we developped Museo ToolBox, an open-source python library that makes spatial cross-validation possible.

作者

我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。

评论

主要评分

4.7
评分不足

次要评分

新颖性
-
重要性
-
科学严谨性
-
评价这篇论文

推荐

暂无数据
暂无数据