Author(s)
Marina Blanton, Nitesh Chawla, Keith Frikken and Troy Raeder
Source
NET Institute Working Paper #08-18, 2008
Summary
This paper provides methods of analysis applicable to understanding consumer patterns of consumption.
Policy Relevance
Retailers can become more profitable by understanding consumer patterns of consumption.
Main Points
- The first problem this paper addresses is how to quickly uncover the important relationships in a large, heterogeneous set of transactions. The most common approach is to find association rules in the data. This paper addresses problems with that approach and provides a procedure for modeling transactional data as a network, methods for analyzing the network, and how the methods apply to real-world data and provide results.
- This paper also develops a protocol for sharing networks of products securely to solve the problem of sample selection bias that exists because individual stores only analyze the data they receive from their store. This biased view of the consumer base can quickly become limiting if the store or company tries to expand and reach out to a broader market.
