Detecting collusive spammers in online review communities
Chang Xu · 2013
In this paper, we first define our research problem as to detect collusive spammers in online review communities. Next we present our current progress on this topic, in which we have spotted anomalies by evaluating 15 behavioral features proposed in the state-of-the-art approaches. Then we propose a novel hybrid classification/clustering method to detect colluders in our dataset based on selected informative features. Experimental results show that our method promisingly improve the performance of traditional classifiers by incorporating clustering for the smoothing. Finally, possible extensions of our current work and challenges in achieving them are discussed as our future directions.