VOTCL及其在交叉销售问题上的应用研究摘要.PDFVIP

ISSNN 111777TP Journal of Computer Research and Development 7( 9) : 153915 7, 2010 VOTCL 周广通 尹义龙 郭心建 董彩玲 ( 250101) ( zhouguangtong@ gmail. com) VOTCL and the Study of Its Application on CrossSelling Problems Zhou Guangtong, Yin Yilong, Guo Xinjian, and Dong Cailing ( S ch ool of Comp ute r S cie nce and Technolog y , S h andong Un iv e r sity , J ina n 250101) Abstract Crossselling is regarded as one of the most promising strategies to make profits. T he authors first describe a typical crossselling model, follow ed by analysis show ing that classimbalance and costsensitivity usually coexist in the data sets collected from this domain. In fact, the central issue in realw orld crossselling applications focuses on the identification of potential crossselling customers. How ever, the performance of customer prediction suffers from the problem that class imbalance and costsensitivity are arising simultaneously. T o address this problem, an effective method called V OT CL is proposed. In the first stage, VOTCL generates a number of balanced training data sets by combining undersampling and oversampling techniques; then a base learner is trained on each of the data set in the second stage; finally, VOT CL obtains the final decisionmaking model by using an optimal threshold based voting scheme. T he effectiveness of V OT CL is validated on the crossselling data set provided by PA KDD 2007 competition w here an AU C value of 0. 603 7 is achieved by using the proposed method. The ensemble model also outperforms a single base learner, w hich to some extent show s the efficacy of the proposed optimal threshold based voting scheme. Key words crossselling; classimbalance; costsensitive; optimal threshold based voting; support vector machine 交叉销售已成为企业盈利的重要手段, 如何解决其数据中普遍同时存在的类 不平衡和代价敏 感问题是准确预测交叉