Easy Visualization of Classification Errors
Classification errors are usually evaluated with complex expert-oriented visualizations. The ●classee● online tool offers simplified user-friendly visualizations, together with the traditional expert-oriented visualizations. It makes it easy to explore your classifier's performance, and communicate your results to end-users:
1. Upload your evaluation data
2. Explore and compare classifiers, datasets and parameters
3. Select the best visualization
4. Download your visualization
(Classee was tested on Chrome and Safari. It may not work with other browsers.)
Video Tutorials
How to use ● classee ●? (coming soon)
How to reuse the visualization components? (comign soon)
Open Source Components
The D3 visualizations of ● classee ● are open source. They can be used in your own projects, with the citation: "Visualization from the Classee Project - classee.project.cwi.nl".
Binary Barchart
User-friendly visualization of errors over a threshold parameter.(binary)
Multiclass Barchart
User-friendly visualization of a confusion matrix.(binary)
Donut Chart
Simple visualization of errors, with no threshold parameter.(binary)
Confusion Matrix
Table with numbers of errors for all possible pairs of classes.(multiclass)
ROC Curve
TP Rate and FP Rate over a threshold parameter, with optional display of variance estimates.(multiclass)
PR Curve
Precision and Recall over a threshold parameter, with optional display of variance estimates.(binary)
Confusion Table
Table with numbers of errors over a threshold parameter.(binary)
Distribution Graph
Distribution of items over a threshold parameter.(binary)