There is method called random forests, which basically construct trees on random slices of data and let them vote on outcome. It is still very simple method and it has better over-fitting performance.