ENBIS-17 in Naples
9 – 14 September 2017; Naples (Italy)
Abstract submission: 21 November 2016 – 10 May 2017
Discovering Communities in Customer Purchase Behavior by Means of Social Network Analytics
12 September 2017, 14:30 – 14:50
- Submitted by
- Jasmien Lismont
- Jasmien Lismont (KU Leuven), Bart Baesens (KU Leuven; University of Southampton), Wilfried Lemahieu (KU Leuven), Jan Vanthienen (KU Leuven)
- Direct marketing is gaining more and more attention in business. Segmentation and personalization of marketing communication allow for increased customer value. In order to derive new insights to advance marketing communication, we create a pseudo-social network of customers of a European retailer. This network is based on customers’ purchasing behavior where customers are connected if they have certain product groups in common. This leads to a bipartite graph with two types of nodes: customers and product groups. Any bipartite graph can be projected onto a unipartite graph with only one type of nodes, i.e. customers directly linked to customers. However, since we are working on a real-life dataset, challenges frequently associated with big data occur. As such, specific applications are required.
Consecutively, descriptive social network techniques are applied to the customer network. Specifically, we apply community mining in order to provide understanding for the development of customer target groups. Various algorithms exist, such as modularity-based, spectral, and dynamic algorithms; but most algorithms are developed for unipartite graphs. Some metrics for bipartite community mining exist which can offer a solution, e.g. BRIM and CoClusLSH. Furthermore, the complexity of these algorithms needs to be taken into account. Many algorithms, e.g. based on betweenness, are not feasible for a large network. Finally, profit measures such as recency, frequency and monetary values, or customer lifetime values can be attached to the customer groups or clusters found in the network. This work is currently still ongoing.
Return to programme