Prediction in economic networks

Vasant Dhar, Tomer Geva, Gal Oestreicher-Singer, Arun Sundararajan

Research output: Contribution to journalArticlepeer-review


We define an economic network as a linked set of entities, where links are created by actual realizations of shared economic outcomes between entities. We analyze the predictive information contained in a specific type of economic network, namely, a product network, where the links between products reflect aggregated information on the preferences of large numbers of individuals to co-purchase pairs of products. The product network therefore reflects a simple "smoothed" model of demand for related products. Using a data set containing more than 70 million observations of a nonstatic co-purchase network over a period of two years, we predict network entities' future demand by augmenting data on their historical demand with data on the demand for their immediate neighbors, in addition to network properties, specifically, local clustering and PageRank. To our knowledge, this is the first study of a large-scale dynamic network that shows that a product network contains useful distributed information for demand prediction. The economic implications of algorithmically predicting demand for large numbers of products are significant.

Original languageEnglish
Pages (from-to)264-284
Number of pages21
JournalInformation Systems Research
Issue number2
StatePublished - Jun 2014


  • Autoregressive models
  • Co-purchase network
  • Economic networks
  • Network-based prediction
  • Neural networks
  • PageRank
  • Prediction
  • Predictive modeling


Dive into the research topics of 'Prediction in economic networks'. Together they form a unique fingerprint.

Cite this