Evaluation

In the Evaluation page, you can compute various statistics and create plots of the results acquired in Modeling. The page consists of two tabs, Statistics and Plotting.

Statistics

In the Statistics tab, you can compute classification and probability metrics for prediction results. First, select deposits (1) and the prediction result rasters (2-3). After selecting the number of reported digits, press Compute to begin. The following metrics are computed:

• Classification metrics
  • Accuracy
  • Precision
  • Recall
  • F1
  • True positives
  • False positives
  • False negatives
  • True negatives
• Probability metrics
  • ROC AUC
  • Log loss
  • Average precision
  • Brief score loss.

Plotting

In the Plotting tab, you can create various plots of the prediction results. First, select plot type (1). The following plot types are available:

• Confusion matrix
• Calibration curve
• DET curve
• Precision-recall curve
• ROC curve Then, select the inputs, i.e. deposits (2) and prediction rasters (3-4). If you selected calibration curve as plot type, you can set the number of bins (5). For all plot types, you can additionally set the resolution of the plot as DPI (dots per inch) (6) and set a save path for the figure (7). To visualize the results, press Create plot.