Hangzhou Xishang Technology Co., Ltd.AI · Software · Data · Simulation
Contact Us
← Back to tutorial library

NEUROBLOCKS STUDIO · Examples & Tutorials

ROC Curve Evaluation

Inspect ROC curves and compare class-wise discrimination performance.

Learn how to use NeuroBlocks Studio for ROC Curve Evaluation with real, manually matched screenshots. This tutorial covers setup, execution, result interpretation and practical checks.

What this tutorial covers

Inspect ROC curves and compare class-wise discrimination performance.

Before you start：You need access to the product and should ideally follow the tutorial using your own real parameters or data. The screenshots below were manually rematched to avoid the image/topic mismatches seen in the previous version.
ROC Curve Evaluation - Open the relevant module and confirm the starting state.
Open the relevant module and confirm the starting state.
ROC Curve Evaluation - Configure the key inputs that match the topic.
Configure the key inputs that match the topic.
ROC Curve Evaluation - Run the workflow and inspect the result/status.
Run the workflow and inspect the result/status.

Step-by-step workflow

1Open the module

After training, switch to the ROC result view. Confirm that the visible page really matches the tutorial topic before changing any settings.

2Configure the input

Prepare and check class scores/probabilities and test labels produced by the completed run.

3Run and watch the status

Inspect per-class ROC curves and compare AUC values. Watch the visible status, plot or log area while the workflow is running.

4Inspect the result

Pay special attention to ROC curves and AUC performance.

5Verify and iterate

ROC is especially useful for revealing weak classes that can be hidden by a good overall accuracy. Change one meaningful parameter and rerun so you can verify whether the output changes in the expected direction.

How to read the result

  • The main output of this workflow is ROC curves and AUC performance.
  • Read the result together with the exact inputs used for the run; a screenshot alone rarely tells the whole story.
  • ROC is especially useful for revealing weak classes that can be hidden by a good overall accuracy.

Troubleshooting & practical tips

  • If nothing changes, first confirm that the correct module is open.
  • If the output looks unrealistic, check units, field mapping and input scale first.
  • For comparisons, change one key parameter at a time whenever possible.
Previous tutorialTraining Hyperparameters & Epoch SetupNext tutorialLoss & Accuracy Curves