Demo: Online Classification on Data Streams

In this example, you have the choice between different algorithms and different datasets. Each dataset consist of a different concept drift type. "IBL-DS" is our instance based algorithm for datastreams and "RBL-DS" is a rule based variant. "Win400" is a default instance base learner which works on a sliding window of the last 400 examples.

The grid of small points represent the true concept of the data generator. Each color is equivalent to a class. The big points are the current instance set of the classifier. In each step 5 new trainings instances are added.

JRE1.5 is necessary!!!