Extract reputable users for trust-aware applications

Abstract

Some trust-aware applications only involve the information provided by reputable users to ensure the reliability. A common way to extract the reputable users is to set a threshold value for the user reputation, and regard those whose reputations bigger than this value as the reputable users. However, existing works just intuitively set the threshold value for the reputable user extraction and did not consider the ratio of available information by setting this value. In this work, we analyze the distribution of the user reputation, and give clear criteria on setting the threshold value for reputable user extraction. Furthermore, we analyze 17 trust network datasets extracted from the real applications to give a general guidance on setting the reputation threshold value. We suggest that for the normalized reputation varying from 0 to 1, the threshold value for the reputable user extraction should be no more than 0.2 to ensure around 10% information available for use.