A novel ensemble pruning algorithm based on randomized greedy selective strategy and ballot

Although the Directed Hill Climbing Ensemble Pruning (DHCEP) algorithm has achieved favorable classification performance, it often yields suboptimal solutions to the ensemble pruning problem, due to its limited exploration within the whole solution space, which inspires us with the development of a novel Ensemble Pruning algorithm based on Randomized Greedy Selective Strategy and Ballot (RGSS&B-EP), where randomization technique is introduced into the procedure of greedy ensemble pruning, and the final pruned ensemble is generated by ballot, which are the two major contributions of this paper. Experimental results, including t-tests on the three benchmark classification tasks, verified the validity of the proposed RGSS&B-EP algorithm.