Development and Evaluation of Mathematical Models for Ubiquitous Data Mining

点击翻译按钮获取中文摘要

The objective of this research was to propose a concept of ubiquitous data mining and develop its mathematical models. Ubiquitous data mining is a new knowledge discovery technology suited for advanced ubiquitous IT environment including Web 2.0. In order to achieve the optimal integration of earlier models, we developed a new algorithm based on the One-Class Support Vector Machine (OC-SVM), and the result of this investigation was presented at the international conference, SAMO2007. The head investigator (M. Koda) was invited to edit the special issue of the Transactions of the Operations Research Society of Japan (in Japanese) on OC-SVM, and complied and published our results in the same issue (vol. 51, No. 11, pp. 677-682, 2006).We further investigated boosting techniques as a means to incorporate multiple data sources including data from ubiquitous sensing devices such as IC tags and RFID (Radio Frequency Identification). As a result, we derived a new robust ensemble learning algorithm, and results were published in the Journal of the Operations Research Society of Japan (vol. 51, pp. 95-110, 2008). The research result on the stochastic sensitivity analysis of financial engineering was also reported at the international conference, ICCS2007, and published in the Lecture Notes in Computer Science (No. 4488, 447-454, 2007). Based on the support vector machine techniques, we explored a customer segmentation methodology and developed a ubiquitous data mining tool for customer relationship management. We prototyped and benchmarked the tool against conventional mining technologies, and the result was published in the European Journal of Operational Research (vol. 186, pp. 358-379, 2008).Jointly with the investigator and graduate students under supervision, we completed a preliminary design of the prototype model for ubiquitous data mining.