Climate Change, Extreme Events and Disaster Risk Reduction by Suraj Mal R. B. Singh & Christian Huggel

Author:Suraj Mal, R. B. Singh & Christian Huggel
Language: eng
Format: epub
Publisher: Springer International Publishing, Cham

Statistical Analysis-Based Evaluations

Receiver operating characteristic curve is usually utilized as a quantitative method to evaluate the general performance of landslide models (Pradhan 2010). Moreover, statistical measures, namely positive predictive value, negative predictive value, positive sensitivity, specificity, and accuracy, are often used to evaluate more detail of the performance of landslide models (Pham et al. 2016b; Tien Bui et al. 2016a). In this study, these methods have been adopted to validate the performance of CART for spatial prediction of landslides.

Receiver operating characteristic (ROC) curve is constructed by plotting pairs of two statistical values, namely “sensitivity” and “100-specificity” (Pham et al. 2016c; Tien Bui et al. 2016b). The value of area under the ROC curve (AUC) is usually utilized to validate quantitatively the performance of landslide models. The AUC value differs from 0.5 to 1.0. As the AUC value is equal to 0.5, the performance of landslide models is inaccurate (Pham et al. 2016d). In contrast, if the AUC value is equal to 1.0, the performance of landslide models is perfect. In general, if the AUC value is higher than 0.8, the performance of landslide models is good and acceptable (Pham et al. 2016e).

Positive predictive value is defined as the probability of pixels that are predicted correctly as “landslide.” Negative predictive value is defined as the probability of pixels that are predicted correctly as “non-landslide.” Sensitivity is the proportion of landslide pixels that are predicted correctly as “landslide.” Specificity is the proportion of non-landslide pixels that are predicted correctly as “non-landslide” (Pham et al. 2016f). Accuracy is the proportion of landslide and non-landslide pixels that are predicted correctly. These statistical measures are calculated based on the values obtained from confusion matrix (Manel et al. 2001). Higher values of these indexes indicate better performance of landslide models.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.