Research Article
An Adaptive Heterogeneous Online Learning Ensemble Classifier for Nonstationary Environments
| | Input: Stream of examples and class labels | | | : set of diverse and accurate learners | | | : Max size of ensemble | | | : global and local predictions | | (1) | For learner = 1 to //loop over learners | | (2) | For j = 1 to m//loop over instances | | (3) | = classify ()//classify with dynamic pool | | (4) | If (i mod = 0) then | | (5) | If () | | (6) | reset | | (7) | recalculate | | (8) | | | (9) | If ( and ) then | | (10) | remove I//delete learner with least Accuracy and Diversity | | (11) | End if Call Active Drift Handler (, ) | | (12) | End for | | (13) | If (i mode = 0) then | | (14) | update accuracy and diversity | | (15) | If (), then Call Passive Drift handler | | (16) | End if | | (17) | If size () = , then | | (18) | (, ) remove ()min | | (19) | End if | | (20) | For i = 1 to | | (21) | Train (, ) | | (22) | End for | | (23) | End if |
|
