2011 15th Panhellenic Conference on Informatics (2011)
Sept. 30, 2011 to Oct. 2, 2011
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/PCI.2011.77
Recommendation systems are at the core of Web 2.0 acting like "virtual" salespersons that collaboratively filter and propose products to consumers of Web stores. This article investigates the Amazon co-purchase network, which is formed by every recommendation of the top selling products in all product categories. Analysis of motifs and communities has been employed in order to understand the basic patterns of aggregate consumption. Network analysis revealed that Amazon has evolved into a book-based multi-store with strong cross-category connections. In addition, co-purchase links not only manifest complementary consumption, but also switching among competitive products. As expected, top selling products are important in the co-purchase network, acting as hubs, authorities and brokers in consumer preferences. Unexpectedly, it was found that ostensibly competitive products are consumed as complements because of the existence of compatibility and compatible products that facilitate their joint consumption.
item-based collaborative filtering; Amazon co-purchase network; subgraph analysis; market basket analysis
D. Kontokostas, T. Theodoridis and M. Vafopoulos, "Inter-viewing the Amazon Web Salespersons: Trends, Complementarities and Competition," 2011 15th Panhellenic Conference on Informatics(PCI), Kastoria, Greece, 2011, pp. 299-303.