Logistic Regression for Detecting Untrustworthy Recommendations in Pervasive Environments

In recent research, the assessment of the trustworthiness of a certain recommender in pervasive environments is examined upon their previous interactions. However, when it is the initial interaction for a certain user, then recommendations cannot be assessed against trustworthiness. One of the approaches is to refer to a previous interaction of one of the users. However, this approach may give good results but it may also lead into wrong recommendations. In this paper, a method for detecting untrustworthiness in pervasive environments is proposed. After digitizing the data attributes, logistic regression is applied. The data attributes used are the recommendations not users information. The proposed method achieved promising results, which are comparable with other research.