Review of the manuscript entitled "Resampling and ensemble techniques for improving ANN-based high streamflow forecast accuracy" by Snieder et al [post]

2020 unpublished
The authors study the effect of resampling techniques, when integrated with ensemble learning frameworks, on the ability of the ANN based regression ensemble learners to improve prediction of high steam flow events. Two case studies are presented, with different temporal resolution and, essentially, hydrologic topology. One individual learner, that is MLP-ANN, is utilized in this study along with two ensemble models (Bagging and Boosting) as well as a randomized set of members (i.e. RWB model).
more » ... Three resampling plans are examined, RUS, ROS, and SMOTER, to serve as the preprocessing re-sampler stage for the ensemble models. a combination of the latter is used with the ensemble models and all configurations are evaluated. C1 HESSD Interactive comment Printer-friendly version Discussion paper
doi:10.5194/hess-2020-430-rc2 fatcat:sg6oe7qk5bbvfn6zmphn2beod4