Spammer group detection based on cascading and clustering of core figures

The problem of collaborative spamming in e-commerce is gradually increasing, and traditional spammer group detection algorithms usually seem cumbersome and time-consuming when dealing with massive user review data. Thus, this research proposes a spammer group detection algorithm based on cascading and clustering of core figures. First, we extract user evaluation features to identify core figures. Then, we use four user interaction features to assess user collusion degree, construct a weighted homogeneous graph by cascading neighbor nodes around core figures, and apply the Louvain weighted clustering algorithm to obtain candidate groups. Finally, we classify candidate groups based on group spam features. Experimental results based on the Amazon reviews dataset demonstrate the algorithm's effectiveness in identifying groups of spammers.