How the Score Works
Step 1: Make Predictions
Our model looks at each image and predicts five types of dry matter.
ŷj = our prediction for biomass component j
Step 2: Compare to the Ground Truth
Each prediction is compared to the true value from the dataset.
Error = (yj − ŷj)²
Step 3: Apply Component Weights
Each error is multiplied by how important that component is.
wj · (yj − ŷj)²
Step 4: Sum Errors and Compare to Variance
Add up all the weighted errors and compare them to the weighted variance.
Score = 1 − SSR / SST
Step 5: Final Score
Our final score is how much of the weighted variation our model explains.
R² = 0.7505 → This means our model explains 75% of the important variation!