Aggregate Two-way Co-Clustering of Ads and User Analysis for Online Advertisements

摘要

Clustering plays an important role in data mining,as it is used by many applications as a preprocessing step fordata analysis. Traditional clustering focuses on groupingsimilar objects, while two-way co-clustering can group dyadicdata (objects as well as their attributes) simultaneously. In thisresearch, we apply two-way co-clustering to the analysis ofonline advertising where both ads and users need to beclustered. The key data that connect ads and users arecontained in the user-ad link matrix, which denotes the adsthat a user has linked. We proposed a three-staged clusteringthat makes use of the three data matrices to enhance clusteringperformance. In addition, an iterative cross co-clusteringalgorithm is also proposed for two-way co-clustering. Theexperiment is performed using the advertisement and userdata from Morgenstern, a financial social website that focuseson the agent of advertisements. The result shows that threestaged clustering provides better performance than traditionalclustering, while iterative co-clustering completes the taskmore efficiently.