Integration of Data Mining Models and Prototyping of CRM Business Model

本研究的目的是整合数据挖掘模型，推导和获得在以前的研究，原型CRM（客户关系管理）的业务模型。为了实现模型的最优集成，在2003年，我们开发了一种新的优化算法的基础上随机灵敏度分析，这一理论研究的结果在SAMO 2004国际研讨会。我们也研究了手机的大量日志数据，并研究了使用移动的手机在线购物的特点。这可能是第一次对Keitai上可能的移动的商务进行研究，其结果发表在日本直复营销学会会刊《直复营销评论》（Vol.3，pp.29 - 41，2004）上。我们与一位获得法国政府Lavoisier基金的法国研究员合作，提出了一种新的基于格理论的关联规则挖掘算法，并在2005年发表。 关于我们 在RIMS Kokyuroku 1351（pp.122 - 133 2004）。我们还探索了一种新的鲁棒增强算法，使用零-一损失函数，结果发表在RIMS Kokyuroku 1351上（pp.106-121，2004）.在2004年，我们进一步研究了boosting技术作为集成多个学习机的一种手段，并导出了一种新的鲁棒boosting算法来对抗错误标记的噪声数据，部分结果发表在日本运筹学会杂志（Vol.47，pp.182 - 196，2004）上。研究结果将成为参与调查的一名研究生的博士论文的主要部分。我们进行了一项由日本直复营销学会支持的研究项目，研究客户细分的方法，并将数据挖掘模型与传统的RFM模型进行了基准测试。2004年在日本直复营销学会年会上发表了基准测试的结果。在吉田和铃木教授的指导下，我们制作了CRM业务的初步模型，并对其性能进行了评估。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