Strategy AtlasStrategy Analysis
StrategyAtlas is a visual analytics system to enable understanding of complex models
by identifying and interpreting different model strategies.
Traditionally, we can find clusters in the data by applying a 2D projection technique (e.g., UMAP).
We propose using the same approach on feature importance values to find model strategies.
Clusters in the StrategyMap indicate model strategies. However, what does each of these clusters mean?
Our proposed visual analytics system StrategyAtlas offers three methods to identify and interpret model strategies:
A gradient heat map enables seeing how the feature contribution values of a single feature are distributed in the projection space.
Interactive density plots enable the analysis of clusters in terms of multiple features at a glance.
The cluster view helps to understand strategy clusters by representing them as a decision tree (separating the cluster from the other data).
To see how StrategyAtlas enables understanding complex machine learning models, check out the demo video:
Or try out the online demo.
If you want to refer to our system, please cite our paper using the following BibTeX entry: