Start Date
14-12-2012 12:00 AM
Description
We define an economic network as a linked set of products, where links are created by realizations of shared outcomes between entities. We analyze the predictive information contained in an increasingly prevalent type of economic network, a “product network” that links the landing pages of goods frequently co-purchased on e-commerce websites. Our data include one million books in 400 categories spanning two years, with over 70 million observations. Using autoregressive and neural-network models, we demonstrate that combining historical demand of a product with that of its neighbors improves demand predictions even as the network changes over time. Furthermore, network properties such as clustering and centrality contribute significantly to predictive accuracy. To our knowledge, this is the first large-scale study showing that a non-static product network contains useful distributed information for demand prediction, and that this information is more effectively exploited by integrating composite structural network properties into one’s predictive models.
Recommended Citation
Dhar, Vasant; Geva, Tomer; Oestreicher-Singer, Gal; and Sundararajan, Arun, "Prediction in Economic Networks: Using the Implicit Gestalt in Product Graphs" (2012). ICIS 2012 Proceedings. 8.
https://aisel.aisnet.org/icis2012/proceedings/KnowledgeManagement/8
Prediction in Economic Networks: Using the Implicit Gestalt in Product Graphs
We define an economic network as a linked set of products, where links are created by realizations of shared outcomes between entities. We analyze the predictive information contained in an increasingly prevalent type of economic network, a “product network” that links the landing pages of goods frequently co-purchased on e-commerce websites. Our data include one million books in 400 categories spanning two years, with over 70 million observations. Using autoregressive and neural-network models, we demonstrate that combining historical demand of a product with that of its neighbors improves demand predictions even as the network changes over time. Furthermore, network properties such as clustering and centrality contribute significantly to predictive accuracy. To our knowledge, this is the first large-scale study showing that a non-static product network contains useful distributed information for demand prediction, and that this information is more effectively exploited by integrating composite structural network properties into one’s predictive models.