基于支持向量机集成的营销风险预警模型研究  

Marketing Crisis Warning Model Based on Support Vector Machine Ensemble

作　　者：冯兰刚 李春播 FENG Lan-gang LI Chun-bo(School of Management Science ＆ Industrial Engineering, Hebei GEO University, Shijiazhuang 050061, China)

机构地区：[1]河北地质大学管理科学与工程学院,河北石家庄050031

出　　处：《南通大学学报（社会科学版）》2017年第1期111-115,共5页Journal of Nantong University:Social Sciences Edition

基　　金：国家自然科学基金(NSFC70772026);河北省首批青年拔尖人才支持计划

摘　　要：建立营销预警机制是企业控制风险的有效途径,利用基于模糊积分的支持向量机集成算法可以构建营销风险预警模型。研究利用Bagging算法能够重新选取训练集增加弱分类器集成差异度的特性来生成个体分类器,最终实现提高模型泛化能力的目的;基于模糊积分算法对各个子分类器进行集成,从而克服了多数投票集成算法未能考虑子支持向量机分类器重要性的缺陷。研究结果表明,基于模糊积分的支持向量机集成算法适合研究具有小样本特征的营销风险预警问题,其效果优于BP神经网络、非集成的SVM和基于最多投票原则集成的SVMS算法。It is an effective measure for companies to control risk by setting up marketing crisis warning mechanisms. In the paper, marketing crisis warning model is constructed based on support vector machine Ensemble of fuzzy integral. Firstly, every support vector classifier is generated by the algorithm of Bagging which can improve the diversities of SVC in order to raise the forecast ability of model. Secondly, in order to overcome the shortness of the majority voting method which fails to consider the importance of sub-classifiers, Support Vector Machine Ensemble based on fuzzy integral algorithm is introduced. Research shows that this method is more appropriate for the analysis of the marketing risk warning with small sample, and the effect is better than BP neural network, SVM of non-integrated, SVM Ensemble based on majonty voting.

关 键 词：支持向量机 风险预警 模糊积分 神经网络 

分 类 号：F272.35[经济管理—企业管理]