基于行为监控和数据挖掘的动态信任模型  

Dynamic trust model based on behavior monitoring and data mining

作　　者：谭兴斌[1] 李刚[2] 李季[1] 

机构地区：[1]重庆大学计算机学院,重庆400044 [2]重庆大学艺术学院,重庆400044

出　　处：《计算机应用研究》2011年第10期3764-3766,共3页Application Research of Computers

基　　金：中央高校基本科研业务费资助项目(CDJZR11180004)

摘　　要：实体之间的信任关系发生在一定的上下文中,其信任值与影响信任值的多个行为属性之间的关系复杂而且多变,很难用一个一成不变的函数去描述。根据软件传感器监测到的历史行为数据和目标信任值,利用lo-gistic回归分析方法和成对分类法对行为属性与信任值之间的关系模式进行自适应的数据挖掘与知识发现,不需要任何的先验知识和主观假设,从而有效解决了信任值计算的动态性和客观性等问题。实验结果表明,与已有模型相比,该模型的平均误差率和计算效率都有所提高,能够快速有效地分类出实体间的信任等级。Trust relationships between entities occur in a certain context.It is difficult to use a static function to describe the complicated and variable relationship between the trust value and the multiple behavioral attributes.According to the historical behavioral data and target trust value gained by monitoring software sensors,the logistic regression analysis and pair classification method were applied to automatically perform data mining and knowledge discovery on relational schema between behavioral attributes and trust value,prior knowledge and subjective assumptions were not included,therefore,effectively solved issues on dynamic and objectivity of trust value calculation.Experimental results show that,compared with the existing model,the new model can improve the MAPE（mean absolute percentage error） and the computational efficiency and classify the trust level between entities effectively.

关 键 词：动态信任 LOGISTIC回归分析 行为监控 数据挖掘 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]