We will demonstrate the use of both RFs and GBRTs in several of the high-impact weather domains described below.
Both RFs and GBRTs provide the ability to measure the importance of each attribute in the dataset, which is called variable importance.
Delphi method was applied to achieve expert's preferences regarding the requirements to be included in a GBRT. Delphi method is a widely used and accepted method for gathering data from respondents within their domain of expertise .
* Preferences regarding the most important assessment elements to be considered in the GBRT; and
* Preferences pertaining the most important assessment criteria to be considered in the GBRT.
The experts have granted high scores to 8 assessment elements and 16 assessment criteria to be included in a GBRT. Through Delphi method, the results allowed the researcher to construct a suggested requirements for a GBRT based on the group consensus.
Delphi method is used to determine the local expert's preferences regarding the most important assessment elements and criteria to be included in the GBRT. It is believe that the results of this study support the use of the Delphi method.
It should be noted that no manual feature engineering was performed; the GBRT and the optimization routines did all of the feature selection and distance weighting on their own.
The best accuracy was achieved with GBRT, using the implementation directly available in R (gbm package) with the mean absolute error (MAE).
Second, GBRT is used to predict the daily solar energy output based on the interpolated model and additional spatiotemporal features.
The best accuracy was achieved with GBRT. The least absolute deviation loss function was used for robustness, and all hyperparameters were optimized on an internal learning set.
The third-place approach also used GBRTs, with the differences coming in the data preprocessing for training.