一种适用于多类问题的神经网络集成模型

A Neural Network Ensemble Model Adapted to Multi-class Problem

提出了一种基于集成学习的适于多类问题的神经网络集成模型,其基本模块由一个OAA方式的二类别分类器和一个互补多类分类器组成.测试表明,该模型在多类问题上比其他经典集成算法有着更高的精度,并且有较少存储空间和计算时间的优势.

A new neural network ensemble model based on ensemble learning model adapted to multi-class problem is proposed. The base components of the proposed model are composed by the union of a binary classifier of OAA and a complementary multi-class classifier. Experimental results show that the model has higher accuracy than other classical ensemble algorithms for multi-class problems, and it has the superiority of less storage space and computation time.