Ensemble methods that combine a committee of machine-learning models, each known as a member or base learner, have gained research interests in the past decade. One interest on ensemble generation involves the multi-objective approach, which attempts to generate both accurate and diverse members that fulfill the theoretical requirements of good ensembles. These methods resolve common difficulties of balancing the trade-off between accuracy and diversity and have been shown to be advantageous over single-objective methods. This study presents an up-to-date survey on multi-objective ensemble generation methods, including widely used diversity measures, member generation, selection, and integration techniques. Challenges and potential applications of multi-objective ensemble generation are also discussed. WIREs Data Mining Knowl Discov 2015, 5:234–245. doi: 10.1002/widm.1158 This article is categorized under: Algorithmic Development textbackslashtextgreater Ensemble Methods