A New Trust Model Based on Social Characteristics and Reputation Mechanisms Using Best Local Prediction Selection Approach

In this paper we propose a new context-aware model for calculation of direct and reputation-based trust. We propose a categorization of intermediaries based on characteristics of social trust relations. This categorization provides wider sources of information to be considered for calculation of new trust value. After converting the reputation values into a time series, it is embedded in a state space using delayed coordinate embedding. Then a technique of prediction of trust value in current time point and in a given context using statistical method of local model selection based on a consistency criterion is proposed. We claim that prediction based on only the recent values of trust will be more accurate and reliable than using the entire past history.