Extracting Trustworthiness Tendencies Using the Frequency Increase Metric

Computational trust systems are currently considered enabler tools for the automation and the general acceptance of global electronic business-tobusiness processes, such as the sourcing and the selection of business partners outside the sphere of relationships of the selector. However, most of the existing trust models use simple statistical techniques to aggregate trust evidences into trustworthiness scores, and do not take context into consideration. In this paper we propose a situation-aware trust model composed of two components: Sinalpha, an aggregator engine that embeds properties of the dynamics of trust; and CF, a technique that extracts failure tendencies of agents from the history of their past events, complementing the value derived from Sinalpha with contextual information. We experimentally compared our trust model with and without the CF technique. The results obtained allow us to conclude that the consideration of context is of vital importance in order to perform more accurate selection decisions.