科研之友
微信
新浪微博
Facebook
分享链接
摘要
To solve cross-regional data mining problem in distributed sensor networks environment, a distributed customer churn prediction model named DCCP based on Multi-Agent was proposed. Firstly, DCCP defined the concept of eigen multi-branches tree named ET, described the mechanism for collecting data from sensor networks based on Multi-Agent, and elaborated the building algorithm of local ET by using Multi-Agent to access datasets on local sensor nodes for generating initial global ET. Then, after analyzing influencing features of the chain business customer consumption behavior in accordance with the different regional economic development level, per capita GDP and other economic factors, DCCP proposed an eigen rule discrepancy algorithm to merge, cut and optimize global ET on ground of the definition of geographical factor, ET rule and ET rule discrepancy. Finally, global data mining in distributed sensor networks was finished by SVM. The experimental results on a chain commercial enterprise in Zhejiang province show that DCCP has a high accuracy for customer churn prediction and can solve these problems which exist in other mining algorithms such as poor efficiency, low security and privacy.
