guojiasheng / husa Goto Github PK
View Code? Open in Web Editor NEWa selective algorithm named hierarchical selection and dynamic update is proposed, which could optimize the parameters of classifier by using multi-thread technique and select the sub sequence set of classifiers base on hierarchical selection and update the information dynamically. It could solve the problem in the past for choosing classifier to ensemble learning inefficiently. In addition, to cut down the time for ensemble vote, a strategy called divide-and-conquer is proposed. The big vote task can be divided recursively into small child task by dichotomy, then running the task parallel and conquer the vote result. Experimental results show that this selective algorithm has made a significant boost, comparing with traditional classification algorithms.