Robust Combination of Model Selection Methods for Prediction

Wei, Xiaoqiao; Yang, Yuhong

doi:10.5705/ss.2010.023

Cited by 4 publications

(1 citation statement)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…where , is the weight of each model ∈ at each time ∈ Ι; note that ∑ , ∞ =1 = 1 for ≥ 1, is a tuning parameter to control the degree of weighting dependence on the predictive performance (Wei and Yang 2011).…”

Section: Structural Similarity Index Measure (Ssim)mentioning

confidence: 99%

Constraining Uncertainties in CMIP5 Projections of Arctic Sea Ice Volume with Observations

Wang

Liu

Shan

et al. 2020

Preprint

View full text Add to dashboard Cite

Abstract. This study proposes adaptive forecasting through exponential re-weighting based on the Structural Similarity Index Measure (AFTER-SSIM) algorithm to evaluate the performance of global climate models from the Coupled Model Intercomparison Project (CMIP5) under different emission scenarios during 2006 to 2018, attempting to reduce the uncertainty among them. The SSIM approach uses a loss function to obtain more information on the spatial distribution between model outputs and observed data, where the genetic algorithm (GA) is used to optimise the parameters of both seasonal cycles and long-term trends of sea ice concentration and sea ice thickness. The re-weighting mechanism of the AFTER-SSIM algorithm guarantees a performance improvement in sea ice volume simulations as new information is added. Finally, the ranked models have been combined to estimate the future Arctic sea ice volume and navigation possibility through the Arctic Northern Sea Route. Results show that the proposed algorithm reduces the uncertainty among models, sea ice volume will continue to shrink in the future, and the open periods for 1A super vessels are likely to reach to five months ranging from August to December in 2030.

show abstract

Section: Structural Similarity Index Measure (Ssim)mentioning

confidence: 99%