Table 4.
Performance of grain yield prediction on testing data, using variable sets determined from LASSO and random forest, as well as all available variables
| Variables | LASSO selected variables | Random forest selected variables | All 172 variables | |||
|---|---|---|---|---|---|---|
| Sample size | r* | RMSE* (g/plot) | r | RMSE (g/plot) | r | RMSE (g/plot) |
| (1) Predictions of SVM model with Gaussian radial basis kernel | ||||||
| All lines | 0.32 | 320.19 | 0.39 | 306.15 | 0.29 | 314.77 |
| NE lines | 0.58 | 326.97 | 0.77 | 254.66 | 0.72 | 284.08 |
| TX lines | 0.21 | 271.44 | 0.36 | 255.51 | 0.57 | 215.92 |
| WB lines | 0.28 | 271.88 | 0.41 | 236.82 | 0.25 | 264.53 |
| OK and SY lines | 0.39 | 201.45 | 0.45 | 191.06 | 0.36 | 193.31 |
| (2) Predictions of ridge regression model | ||||||
| All lines | 0.49 | 283.86 | 0.39 | 301.89 | 0.25 | 314.83 |
| NE lines | 0.73 | 272.72 | 0.81 | 225.45 | 0.73 | 295.92 |
| TX lines | 0.55 | 235.37 | 0.50 | 255.68 | 0.47 | 242.99 |
| WB lines | 0.40 | 247.57 | 0.42 | 246.21 | 0.22 | 266.25 |
| OK and SY lines | 0.59 | 163.69 | 0.54 | 164.90 | 0.58 | 169.29 |
* Values of r and RMSE were averaged from 20 random sets of testing data