Abstract | ||
---|---|---|
In real time advertising we are interested in finding features that improve click-through rate prediction. One source of available information is the bipartite graph of websites previously engaged by identifiable users. In this work, we investigate three different decompositions of such a graph with varying degrees of sparsity in the representations. The decompositions that we consider are SVD, NMF, and IRM. To quantify the utility, we measure the performances of these representations when used as features in a sparse logistic regression model for click-through rate prediction. We recommend the IRM bipartite clustering features as they provide the most compact representation of browsing patterns and yield the best performance. |
Year | DOI | Venue |
---|---|---|
2014 | 10.1109/MLSP.2014.6958852 | Machine Learning for Signal Processing |
Keywords | Field | DocType |
Bayes methods,Web sites,advertising,feature extraction,graph theory,information retrieval,pattern clustering,regression analysis,singular value decomposition,Bayesian generative model,IRM bipartite clustering features,NMF,SVD,URL,Website,bipartite graph,browsing pattern representation,click-through rate prediction,compact Web browsing profile,graph decomposition,infinite relational model,nonnegative matrix factorization,real time advertising,representation sparsity,singular value decomposition,sparse logistic regression model | Data mining,Computer science,Artificial intelligence,Cluster analysis,Logistic regression,Graph,Singular value decomposition,Click-through rate,Pattern recognition,Bipartite graph,Web navigation,Non-negative matrix factorization,Machine learning | Conference |
ISSN | Citations | PageRank |
2161-0363 | 0 | 0.34 |
References | Authors | |
9 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Bjarne Ørum Fruergaard | 1 | 0 | 0.34 |
Lars Kai Hansen | 2 | 2776 | 341.03 |