Contextual Spatial Outlier Detection with Metric Learning
Contextual Spatial Outlier Detection with Metric Learning
复制标题
DOI：
10.1145/3097983.3098143
复制
发表时间：
2017-08
期刊：
影响因子：
--
通讯作者：
Guanjie Zheng;S. Brantley;T. Lauvaux;Z. Li
中科院分区：
文献类型：
--
作者：
Guanjie Zheng;S. Brantley;T. Lauvaux;Z. Li
水力压裂是一种革命性的页岩气增产技术，但在环境污染方面引发了争议。如果天然气井的甲烷大量泄漏，这种温室气体会影响饮用水井，加剧全球变暖。我们的工作是受到这种对环境问题的激烈辩论的推动，并将重点放在检测异常空间数据样本（例如与潜在泄漏相关的水样）的一般数据分析技术上。具体来说，我们提出了一种基于上下文邻居的空间离群点检测方法。与现有工作不同的是，我们的方法同时利用空间属性和非空间上下文属性来定义邻居。我们进一步使用鲁棒度量学习来组合不同的上下文属性，以找到有意义的邻居。我们的技术可以应用于任何空间数据集。在五个真实数据集上的大量实验结果证明了我们方法的有效性。我们还展示了一些有趣的案例研究，包括一个与气井泄漏有关的案例。
Hydraulic fracturing (or "fracking") is a revolutionary well stimulation technique for shale gas extraction, but has spawned controversy in environmental contamination. If methane from gas wells leaks extensively, this greenhouse gas can impact drinking water wells and enhance global warming. Our work is motivated by this heated debate on environmental issue and focuses on general data analytical techniques to detect anomalous spatial data samples (e.g., water samples related to potential leakages). Specifically, we propose a spatial outlier detection method based on contextual neighbors. Different from existing work, our approach utilizes both spatial attributes and non-spatial contextual attributes to define neighbors. We further use robust metric learning to combine different contextual attributes in order to find meaningful neighbors. Our technique can be applied to any spatial dataset. Extensive experimental results on five real-world datasets demonstrate the effectiveness of our approach. We also show some interesting case studies, including one case linking to leakage of a gas well.