Heterogeneous Data Mining in Search Advertisement Click Rates

Search advertising is typical heterogeneous multi-dimensional data sets whose click-through rate depends on query words, terms and other factors. Traditional data mining methods, limited to homogenous data source, represent search ads as the vector space model, so they fail to sufficiently consider the search advertisements´ characteristics of heterogeneous data. This paper presents consistent bipartite graph model to describe ads, adopting spectral co-clustering method in data mining. In order to solve the balance partition of the map in clustering, heuristic algorithm is introduced into consistent bipartite graph´s co-partition; a more effective subgraph redistribution algorithm is established. Experiments on real ads dataset shows that our approach worked effectively and efficiently in both clustering and prediction.