Integration of Data Mining Models and Prototyping of CRM Business Model

本研究的目的是整合先前研究中所得到的数据挖掘模型，以及客户关系管理(Customer Relationship Management，CRM)业务模型的原型。为了实现模型的最优集成，2003年，我们提出了一种新的基于随机灵敏度分析的优化算法，并在SAMO2004国际研讨会上公布了这一理论研究的结果。我们还研究了科泰手机的海量日志数据，并调查了使用手机进行网上购物的特点。这可能是第一次对Keitai上可能的移动商务进行研究，研究结果发表在《日本直销学术学会期刊》上，《直销评论》(第三卷，第29-41页，2004年)。我们与一位获得法国政府拉瓦锡资助的法国研究人员一起，基于格理论推导出了一种新的挖掘关联规则的算法，并发表在I…上《Kokyuroku 1351》(第122-133页，2004年)。我们还探索了一种新的基于零一损失函数的稳健Boosting算法，其结果发表在RIMS Kokyuroku 1351(pp.106-121,2004)上。2004年，我们进一步研究了Boosting技术作为一种整合多个学习机的手段，并推导出了一种新的针对错误标签噪声数据的稳健Boosting算法，部分结果发表在日本运筹学学会期刊(Vol.47，pp.182-196,2004)上。这一结果将成为一名参与调查的研究生博士论文的主要部分。我们进行了一个由日本直销学会支持的关于客户细分方法的研究项目，并将数据挖掘模型与传统的RFM模型进行了基准比较。标杆管理的结果在2004年日本直销学术学会年会上公布。我们与吉田教授和铃木教授等共同研究人员共同构建了一个初步的CRM业务模型，并对其绩效进行了评估。较少

The objective of this research was to integrate data mining models that were derived and obtained in the previous study, and prototype CRM(Customer Relationship Management) business model. In order to achieve the optimal integration of models, in 2003, we developed a new optimization algorithm based on stochastic sensitivity analysis, and the result of this theoretical investigation was presented at International Symposium SAMO2004. We also studied the huge volume of log data of Keitai-cellular phone, and investigated the characteristics of on-line shopping using mobile phone. This might be the first study of possible mobile commerce on Keitai, and the result was published in the Journal of the Japan Academic Society of Direct Marketing, Direct Marketing Review (vol.3,pp.29-41,2004). Together with a French research fellow who won the Lavoisier grant from French government, we derived a new mining algorithm to discover association rules based on the lattice theory and it was published i … More n the RIMS Kokyuroku 1351 (pp.122-133 2004). We also explored a new robust boosting algorithm using the zero-one loss function, and the result was published in the RIMS Kokyuroku 1351 (pp.106-121,2004).In 2004,we further investigated boosting techniques as a means to integrate multiple learning machines and derived a new robust boosting algorithm against mislabeled noisy data, and partial results were published in the Journal of the Operations Research Society of Japan (vol.47,pp.182-196,2004). The result will become a major part of the doctoral dissertation of a graduate student who collaborated in the investigation.We conducted a research project supported by the Japan Academic Society of Direct Marketing on the methodology of customer segmentations, and benchmarked the data mining models against conventional RFM models. The result of benchmarking was presented at the annual conference of the Japan Academic Society of Direct Marketing in 2004.Jointly with co-investigators, i.e., Professors Yoshida and Suzuki, we prototyped a preliminary model for CRM business, and evaluated its performance. Less