Product Purchasing Networks for Overlapping Community Detection
Disciplines
Applied Statistics | Other Mathematics | Sales and Merchandising
Abstract (300 words maximum)
Research has been done in creating product purchase networks from transaction data and detecting communities of related products to reduce the number of association rules that must be investigated in market basket analysis. These networks use nodes to represents products with edges connecting nodes if the products they represent appear together in a transaction. The edges are weighted with a chosen metric concerning the importance of the rule. The same product may be related to multiple communities for different reasons, and algorithms have been developed to detect these overlapping communities in graphs. This project explores how the use of directed or undirected edges and the metric used to weight the edges in the product purchase network affect the resulting communities.
Academic department under which the project should be listed
CCSE - Data Science and Analytics
Primary Investigator (PI) Name
Joe DeMaio
Product Purchasing Networks for Overlapping Community Detection
Research has been done in creating product purchase networks from transaction data and detecting communities of related products to reduce the number of association rules that must be investigated in market basket analysis. These networks use nodes to represents products with edges connecting nodes if the products they represent appear together in a transaction. The edges are weighted with a chosen metric concerning the importance of the rule. The same product may be related to multiple communities for different reasons, and algorithms have been developed to detect these overlapping communities in graphs. This project explores how the use of directed or undirected edges and the metric used to weight the edges in the product purchase network affect the resulting communities.