|
Abstract:
|
Data on large consumer networks are becoming increasingly available. For example, e-commerce and telecommunications firms are in the business of collecting data on their large proprietary consumer networks and portal sites such as MySpace and Facebook provide publicly available data on links between networked consumers to members of their network. As these consumer network data continue to become larger, firms will need better techniques to use network features for business decision-making. This research adapts collective inference models applied previously in other domains such as image processing and web page classification, to identify likely adopters of a large telecommunication firm's new technology service. We show that our collective inference method, a global approach, improves consumer ranking over existing methods that use only local network features.
|
