LightGBM with surrounding-grid feature selection and Tweedie loss for precipitation yields lower RMSE than raw MSM forecasts, MSMG, and some CNN baselines across many Japanese locations and lead times.
Why do tree-based models still outperform deep learning on typical tabular data? Advances in neural information processing systems, 35: 0 507--520, 2022
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
physics.ao-ph 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Improvements to the post-processing of weather forecasts using machine learning and feature selection
LightGBM with surrounding-grid feature selection and Tweedie loss for precipitation yields lower RMSE than raw MSM forecasts, MSMG, and some CNN baselines across many Japanese locations and lead times.