It seems we can build a dbscan.create() model on training data but there is no way to classify test data into the precomputed clusters. It will be nice to have a classify() method, that for each point in the test data returns either the closest core cluster, or a boundary condition where the point is between core clusters, or a noise indication that the point is anomalous not near any core cluster.
At the very least DBSCANModel should have a predict method (like KmeansModel). Without a predict method, the usefulness of our dbscan toolkit seems very limited.
It seems we can build a dbscan.create() model on training data but there is no way to classify test data into the precomputed clusters. It will be nice to have a classify() method, that for each point in the test data returns either the closest core cluster, or a boundary condition where the point is between core clusters, or a noise indication that the point is anomalous not near any core cluster.