In this paper we analyze the performance of state-of-the-art machine learning techniques in trust prediction. We use two propositionalization methods together with the Naive and Tree-Augmented Naive Bayesian Classifiers, and the C4.5 algorithm. We compare those results with classifiers defined through Markov Logic, using data fromthe Epinions.comwebsite, a well-known product review community. The experiments show that predicting trust relationships is a difficult task, in which Markov Logic models outperform other methods in accuracy but are able to recover only a relatively small fraction of the existing relationships in the dataset.