Dynamic ensemble selection with local expertise consistency

In classification tasks, ensemble selection methods select some base learners from the learners pool instead all of them to classify a query patterns. Static ensemble selection schemes determine the final ensemble immediately after training and apply it to all test patterns. On the other hand, dynamic ensemble selection (DES) construct a customized ensemble for every query pattern by incorporating its local information. Most DES differ each other only on the selection scheme. We propose Dynamic Ensemble Selection with Local Expertise Consistency (DES-LEC) that focus on generating a learners pool dedicated to the latter selection phase. Experiment results on 4 medical data sets suggest that DES-LEC is able to improve the performance over the DES systems that select from a regular learners pool.