M AS中一种基于稳定性的直接信任计算方法

摘要

To evaluate direct trust between Agents more effectively interaction records have to be firstly divided into recent records and history records.The update weight of new interaction is then determined according to three factorsstability of all records abnormal behavior records and difference between the newest record and recent trust which can show the dynamic changes of Agent's recent trust.Meanwhile direct trust is composed of the weights of recent trust and history trust which are decided by the differences between them.The experiments indicate that this method can be used to improve the stability of direct trust and applicable to real situations.%为了更有效地对Agent之间的直接信任度进行评估，首先将交互记录分为近期和历史两部分，然后从交互记录的稳定性、异常行为记录以及新旧交互记录之间的差异三个方面的因素决定新交互记录的更新权重，从而动态地反映出近期信任度的变化；同时由近期信任度与历史信任度根据两者之间的差异决定各自的权重，共同构成Agent的直接信任度。试验证明该方法有效提高了直接信任度的可靠性，更适用于实际环境。