【作者】 李征；

【导师】 杨文川；

【作者基本信息】 北京邮电大学 ， 软件工程， 2008， 硕士

【摘要】 随着电信体制改革的深化和加入WTO,我国固话通信运营业的竞争也日趋激烈。与其他行业相比,固话通信运营业拥有更多有关用户的数据。谁能正确地挖掘与分析隐含于这些数据中的知识,谁就能更好地向用户提供产品与服务。能够发现更多的商机,从而在竞争中获胜。国内对这方面的研究还处于起步阶段,国外在这方面已经大大地超前于国内。因此,数据挖掘在我国固话通信运营业中的研究有重要应用价值。本文重点描述了北京网通公司如何开展及运用数据挖掘技术来提高其竞争力,本文没有对数据挖掘理论及建模方法等做过多的阐述,也没有对数据仓库建设方面做过细的探讨。而是将重点放在数据挖掘模型的选择和设计上,在国外已有研究的基础上,结合北京网通公司实际需要,选取现网数据分别采用决策树和回归模型在SASEnterprise Miner上验证和评价。本论文的主要内容是:1.对北京网通客户流失预测的现状进行了阐述,确定了研究的对象,研究方法以及研究需要所要达到的效果。2.讲述了如何针对高价值客户分析、建立流失预测模型。3.采用决策树和回归方法得出流失预测模型,并进行对比。4.流失预测系统在北京网通的实际应用效果。更多还原

【Abstract】 With the telecommunication system reform development and our entrance of WTO, the competition of telecommunication operation is becoming fiercer. Compared with other industry, the mobile telecommunication operation has more data of customers. Who can mine and analyze the knowledge contained in the data correctly will offer product and service to customer better and find more opportunities, thus win the competition. Our domestic research on this field is still at infancy, companies abroad have already been superior to mine greatly. So, there is important practical value in the research on data mining of our telecommunication industry.This thesis mainly researches on our mobile telecommunication operation how to launch and use data mining to raise its competitive advantage. This thesis has not discussed much on data mining theory and the modeling method, etc. Nor has the construction of data warehouse. We put focal point on the choice and design that the model of data mining, on the basis of the already studies of abroad and the actual needs of Beijing China net communication company. Put real data on Decision Trees or regression model, and SAS Enterprise Miner was used to test and appraise these models.The main contribution of this thesis are follows:1. Described the actuality of high-value customers losing predication of China Netcom. Set up the research object, method and effect to be achieved.2. Described how to analyze and make the predication model for high-value customers losing. 3. Introduced Decision Trees and Regression methods to obtain two predication models, and compared them.4. The actual effects of custom losing prediction system in China Netcom.更多还原