一种基于实体上下文和时间戳的信任预测模型

信任是网络计算模式下实体交互与协同的基础，如何准确地定量表示和评估信任是研究重点。该文提出以实体上下文和时间戳为条件的信任预测模型，建立了粒度为8的信任等级空间，引入了多维测量指标度量实体交互满意度，使得满意度计算更加精确。构建了具有时间衰减性的直接信任求解方法，克服了已有模型动态适应能力不足的问题。把推荐信任划分为直接推荐和间接推荐，在直接推荐信任求解中引入实体评分相似度因子，在间接推荐信任计算中提出了基于路径衰减的方法。提出了一种分布式树型存储机制DST(Distributed Storage Tree)，提高了模型的稳定性和可扩展性。模拟实验表明，与已有同类型模型相比，该模型更有效和准确地提供决策依据，并且在抑制恶意实体方面具有明显作用。

A Trust Forecasting Model Based on Entity Context and Time Stamp

In network computing model, trust is the major driving force for interaction and collaboration among distributed entities. How to accurately quantify and evaluate trust is the research focus. A trust forecasting model is proposed based on entity context and time stamp. Trust grade space with 8 granularity is established. The multi-dimensional measurement standard is introduced to more scientific and accurate measure of interaction satisfaction degree. The direct trust calculation method with time self-decay is proposed, which overcomes the deficiency of existing models that do not adapt to dynamic changes environment. The recommendation trust is divided to direct and indirect recommendation trust. A similarity factor of entity scoring activities is introduced to caculate direct recommendation trust. The path self-decay factor is given to calculate indirect recommendation trust. A Distributed Storage Tree (DST) mechanism is proposed, which can improve stability and scalability of model. Simulation results show that compared to the exisitng trust model, the model can more effectively and accurately provide decision-making basis, and effectively inhibit malicious entities.