Product network analysis

There are two steps in product network analysis. The first step is to transform the point-of-sale data into product pairs and their transaction frequency. The second step is to create a graph using data from the first step and run a clustering algorithm on the graph. The subgraphs or the clusters formed are presented as the micro-categories. Also, some products in the graph play key roles. Clustering and visualizing these product subgraphs will also help us identify those key products. According to the white paper by Corte Consultancy, a product fitting any of the following definitions is considered as key to the network:

• The core product: In a subgraph or a cluster group, the product that is most commonly purchased in the group is termed as the core product of that group.
• The connectors: These are products that connect two subgraphs or clusters together. They are the ones that are typically bought first, if a customer starts shopping for products in that subgraph...