Automatic seed set expansion for trust propagation based anti-spamming algorithms
Xianchao Zhang, Bo Han, Wenxin Liang · 2009
Seed sets are of significant importance for trust propagation based anti-spamming algorithms, e.g., TrustRank. Conventional approaches require manual evaluation to construct a seed set, which restricts the seed set to be small in size, since it would cost too much and may even be impossible to construct a very large seed set manually. The small-sized seed set can cause detrimental effect on the final ranking results. Thus, it is desirable to automatically expand an initial seed set to a much larger one. In this paper, we propose the first automatic seed set expansion algorithm (ASE), which expands a small seed set by selecting reputable seeds that are found and guaranteed to be reputable through a joint recommendation link structure. Experimental results on the WEBSPAM-2007 dataset show that with the same manual evaluation efforts, ASE can automatically obtain a large number of reputable seeds with high precision, thus significantly improving the performance of the baseline algorithm in terms of both reputable site promotion and spam site demotion.