Table 2.
Cross-Validated R2 for Different Regions
Ensemble Model | Neural Network | Random Forest | Gradient Boosting | ||||||
---|---|---|---|---|---|---|---|---|---|
Region | R2 | RMSE | Bias | Slope | Spatial R2 | Temporal R2 | R2 | R2 | R2 |
East North Central | 0.924 | 2.106 | 0.344 | 0.981 | 0.915 | 0.925 | 0.917 | 0.921[a] | 0.899 |
East South Central | 0.894 | 2.214 | 0.533 | 0.964 | 0.897 | 0.891 | 0.888 | 0.891[a] | 0.872 |
Middle Atlantic | 0.893 | 2.512 | 0.420 | 0.978 | 0.870 | 0.900 | 0.887[a] | 0.883 | 0.863 |
Mountain | 0.769 | 3.337 | 0.891 | 0.911 | 0.769 | 0.764 | 0.765 | 0.769[a] | 0.694 |
New England | 0.889 | 2.122 | 0.356 | 0.979 | 0.878 | 0.894 | 0.883[a] | 0.875 | 0.852 |
Pacific | 0.802 | 4.045 | 0.929 | 0.936 | 0.850 | 0.776 | 0.797[a] | 0.797 | 0.750 |
South Atlantic | 0.895 | 2.117 | 0.441 | 0.974 | 0.907 | 0.890 | 0.888 | 0.893[a] | 0.872 |
West North Central | 0.862 | 2.326 | 0.408 | 0.966 | 0.875 | 0.854 | 0.856[a] | 0.851 | 0.824 |
West South Central | 0.850 | 2.246 | 0.717 | 0.942 | 0.788 | 0.857 | 0.841[a] | 0.841 | 0.824 |
the learner outperformed other learners in that location.
Note: Region division was based on U.S. Census Bureau. New England: Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont; Middle Atlantic: New Jersey, New York, Pennsylvania; East North Central: Indiana, Illinois, Michigan, Ohio, Wisconsin; West North Central: Iowa, Nebraska, Kansas, North Dakota, Minnesota, South Dakota, Missouri; South Atlantic: Delaware, District of Columbia, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia, West Virginia; East South Central: Alabama, Kentucky, Mississippi, Tennessee; West South Central: Arkansas, Louisiana, Oklahoma, Texas; Mountain: Arizona, Colorado, Idaho, New Mexico, Montana, Utah, Nevada, Wyoming; Pacific: Alaska, California, Hawaii, Oregon, Washington. Although the Pacific Region includes Alaska and Hawaii, both states were not included in our modeling.