Table 2. Characteristics of Classification Models and Correctly Predicted Crude Test Set Samplesa.
| OPLS-DA |
RF |
||||||||
|---|---|---|---|---|---|---|---|---|---|
| classification model | attribution capacity | class | comp.a | R2X (cum) | R2Y (cum) | Q2(cum) | prediction | OOB error (%) | prediction |
| M1crude | ethylene or TDG routes | R(1–9) | 1 + 2 + 0 | 0.72 | 0.99 | 0.95 | 9/9 | 2.3 | 9/9 |
| R(10, 11) | 2/2 | 2/2 | |||||||
| M2crude | chlorination methods | R(1, 4, 7) | 2 + 2 + 0 | 0.61 | 0.98 | 0.95 | 3/3 | 3/3 | |
| R(2, 5, 8) | 3/3 | 0.0 | 3/3 | ||||||
| R(3, 6, 9) | 3/3 | 3/3 | |||||||
| M3a crude | TDG synthesis methods of R(1, 4, 7) samples | R1 | 2 + 2 + 0 | 0.70 | 0.98 | 0.90 | 1/1 | 1/1 | |
| R4 | 0/1 | 33.3 | 1/1 | ||||||
| R7 | 0/1 | 1/1 | |||||||
| M3bcrude | TDG synthesis methods of R(2, 5, 8) samples | R2 | 2 + 1 + 0 | 0.63 | 0.98 | 0.88 | 1/1 | 1/1 | |
| R5 | 1/1 | 16.7 | 1/1 | ||||||
| R8 | 0/1 | 1/1 | |||||||
| M3ccrude | TDG synthesis methods of R(3, 6, 9) samples | R3 | 2 + 1 + 0 | 0.55 | 0.97 | 0.91 | 1/1 | 1/1 | |
| R6 | 0/1 | 8.3 | 0/1 | ||||||
| R9 | 1/1 | 1/1 | |||||||
a
Comp. shows the number of components (x/y joint predictive variation + variation in x orthogonal to y + variation in y orthogonal to x) included in each model.