排序方式: 共有13条查询结果,搜索用时 15 毫秒
11.
Jonne Pohjankukka Tapio Pahikkala Paavo Nevalainen Jukka Heikkonen 《International journal of geographical information science》2017,31(10):2001-2019
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. 相似文献
12.
This paper analyses the experiences acquired through several research projects on the history of water and sanitation services by two multidisciplinary teams. Challenges have been faced in identifying feasible objectives, realistic resources, time allocations and unexpected external factors. Water history can preserve cultural heritage, promote reputation management, record vanishing knowledge, and discover new facts. 相似文献
13.
Evaluation of erosion and surface roughness in peatland forest ditches using pin meter measurements and terrestrial laser scanning 下载免费PDF全文
Leena Stenberg Tapio Tuukkanen Leena Finér Hannu Marttila Sirpa Piirainen Bjørn Kløve Harri Koivusalo 《地球表面变化过程与地形》2016,41(10):1299-1311
Anthropogenic activities on peatlands, such as drainage, can increase sediment transport and deposition downstream resulting in harmful ecological impacts. The objective of this study was to quantify changes in erosion/deposition quantities and surface roughness in peatland forest ditches by measuring changes in ditch cross‐sections and surface microtopography with two alternative methods: manual pin meter and terrestrial laser scanning (TSL). The methods were applied to a peat ditch and a ditch with a thin peat layer overlaying erosion sensitive mineral soil within a period of two years following ditch cleaning. The results showed that erosion was greater in the ditch with exposed mineral soil than in the peat ditch. The two methods revealed rather similar estimates of erosion and deposition for the ditch with the thin peat layer where cross‐sectional changes were large, whereas the results for smaller scale erosion and deposition at the peat ditch differed. The TLS‐based erosion and deposition quantities depended on the size of the sampling window used in the estimations. Surface roughness was smaller when calculated from the pin meter data than from the TLS data. Both methods indicated that roughness increased in the banks of the ditch with a thin peat layer. TLS data showed increased roughness also in the peat ditch. The increase in surface roughness was attributed to erosion and growth of vegetation. Both methods were suitable for the measurements of surface roughness and microtopography at the ditch cross‐section scale, but the applicability, rigour, and ease of acquisition of TLS data were more evident. The main disadvantage of the TLS instrument (Leica ScanStation 2) compared with pin meter was that even a shallow layer of humic (dark brown) water prevented detection of the ditch bed. The geomorphological potential of the methods was shown to be limited to detection of surface elevation changes >~0.1 m. Copyright © 2016 John Wiley & Sons, Ltd. 相似文献