Table 4.

Mean and standard deviation of different performance metrics across 10 repetitions of a three-fold cross-validation for different parameter configurations in the 2cats and 5cats direct classifications. Precision, recall, and AUC-ROC values correspond to the macro average. Note that for a random classification with a uniform class distribution the accuracy, precision, and recall are all at 50 per cent for 2cats and 20 per cent for 5cats, while AUC-ROC is at 50 per cent in both tasks.

ExperimentPerformance metrics
ClassificationInput dataAccuracyPrecisionRecallAUC-ROC
2catsColour|$0.821\pm 0.006$||$0.852\pm 0.009$||$0.853\pm 0.009$||$0.893\pm 0.005$|
 Structural1|$0.825\pm 0.006$||$0.855\pm 0.011$||$0.858\pm 0.007$||$0.904\pm 0.005$|
 Structural2|$0.850\pm 0.006$||$0.875\pm 0.009$||$0.880\pm 0.009$||$0.923\pm 0.004$|
 S2+C|$0.869\pm 0.007$||$0.892\pm 0.008$||$0.893\pm 0.009$||$0.943\pm 0.004$|
5catsColour|$0.520\pm 0.008$||$0.490\pm 0.014$||$0.455\pm 0.009$||$0.829\pm 0.004$|
 Structural1|$0.541\pm 0.009$||$0.489\pm 0.035$||$0.454\pm 0.008$||$0.831\pm 0.004$|
 Structural2|$0.589\pm 0.008$||$0.534\pm 0.026$||$0.502\pm 0.007$||$0.860\pm 0.004$|
 S2+C|$0.634\pm 0.007$||$0.608\pm 0.012$||$0.580\pm 0.011$||$0.897\pm 0.003$|
ExperimentPerformance metrics
ClassificationInput dataAccuracyPrecisionRecallAUC-ROC
2catsColour|$0.821\pm 0.006$||$0.852\pm 0.009$||$0.853\pm 0.009$||$0.893\pm 0.005$|
 Structural1|$0.825\pm 0.006$||$0.855\pm 0.011$||$0.858\pm 0.007$||$0.904\pm 0.005$|
 Structural2|$0.850\pm 0.006$||$0.875\pm 0.009$||$0.880\pm 0.009$||$0.923\pm 0.004$|
 S2+C|$0.869\pm 0.007$||$0.892\pm 0.008$||$0.893\pm 0.009$||$0.943\pm 0.004$|
5catsColour|$0.520\pm 0.008$||$0.490\pm 0.014$||$0.455\pm 0.009$||$0.829\pm 0.004$|
 Structural1|$0.541\pm 0.009$||$0.489\pm 0.035$||$0.454\pm 0.008$||$0.831\pm 0.004$|
 Structural2|$0.589\pm 0.008$||$0.534\pm 0.026$||$0.502\pm 0.007$||$0.860\pm 0.004$|
 S2+C|$0.634\pm 0.007$||$0.608\pm 0.012$||$0.580\pm 0.011$||$0.897\pm 0.003$|
Table 4.

Mean and standard deviation of different performance metrics across 10 repetitions of a three-fold cross-validation for different parameter configurations in the 2cats and 5cats direct classifications. Precision, recall, and AUC-ROC values correspond to the macro average. Note that for a random classification with a uniform class distribution the accuracy, precision, and recall are all at 50 per cent for 2cats and 20 per cent for 5cats, while AUC-ROC is at 50 per cent in both tasks.

ExperimentPerformance metrics
ClassificationInput dataAccuracyPrecisionRecallAUC-ROC
2catsColour|$0.821\pm 0.006$||$0.852\pm 0.009$||$0.853\pm 0.009$||$0.893\pm 0.005$|
 Structural1|$0.825\pm 0.006$||$0.855\pm 0.011$||$0.858\pm 0.007$||$0.904\pm 0.005$|
 Structural2|$0.850\pm 0.006$||$0.875\pm 0.009$||$0.880\pm 0.009$||$0.923\pm 0.004$|
 S2+C|$0.869\pm 0.007$||$0.892\pm 0.008$||$0.893\pm 0.009$||$0.943\pm 0.004$|
5catsColour|$0.520\pm 0.008$||$0.490\pm 0.014$||$0.455\pm 0.009$||$0.829\pm 0.004$|
 Structural1|$0.541\pm 0.009$||$0.489\pm 0.035$||$0.454\pm 0.008$||$0.831\pm 0.004$|
 Structural2|$0.589\pm 0.008$||$0.534\pm 0.026$||$0.502\pm 0.007$||$0.860\pm 0.004$|
 S2+C|$0.634\pm 0.007$||$0.608\pm 0.012$||$0.580\pm 0.011$||$0.897\pm 0.003$|
ExperimentPerformance metrics
ClassificationInput dataAccuracyPrecisionRecallAUC-ROC
2catsColour|$0.821\pm 0.006$||$0.852\pm 0.009$||$0.853\pm 0.009$||$0.893\pm 0.005$|
 Structural1|$0.825\pm 0.006$||$0.855\pm 0.011$||$0.858\pm 0.007$||$0.904\pm 0.005$|
 Structural2|$0.850\pm 0.006$||$0.875\pm 0.009$||$0.880\pm 0.009$||$0.923\pm 0.004$|
 S2+C|$0.869\pm 0.007$||$0.892\pm 0.008$||$0.893\pm 0.009$||$0.943\pm 0.004$|
5catsColour|$0.520\pm 0.008$||$0.490\pm 0.014$||$0.455\pm 0.009$||$0.829\pm 0.004$|
 Structural1|$0.541\pm 0.009$||$0.489\pm 0.035$||$0.454\pm 0.008$||$0.831\pm 0.004$|
 Structural2|$0.589\pm 0.008$||$0.534\pm 0.026$||$0.502\pm 0.007$||$0.860\pm 0.004$|
 S2+C|$0.634\pm 0.007$||$0.608\pm 0.012$||$0.580\pm 0.011$||$0.897\pm 0.003$|
Close
This Feature Is Available To Subscribers Only

Sign In or Create an Account

Close

This PDF is available to Subscribers Only

View Article Abstract & Purchase Options

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.

Close