将多分类器融合技术用于CRM中的客户分类研究,以提高分类性能。以决策树作为基本分类器,引入最小二乘技术进行多分类器线性融合。实证结果显示,4种不同的融合方案的分类性能均胜过任一基本分类器,甚至优于基于遗传算法的神经网络融合分类结果,从而表明了该方法的可行性和有效性。
Applying multi -classifiers fusion technology to customer classification in CRM to improve the classification performance. This paper regards decision tree as basic classifier, and introduces the least square technology for linear fusion. It shows that, the classification performances of four different fusion schemes all outperform any basic classifier, even outperform the fusion of neural networks based on genetic algorithm, and thereby demonstrate the feasibility and validity of this method.