Use of Clusterization Technique to Highlight Groups of Related Goods by Digital Traces in Retail Trade

One of the most important management technologies for the development of trade business is the category management directed at distinguishing groups of complementary and substitute products with reference to the analysis of shopping preferences. The traditional method based on a customer survey conducted at the supermarket exits is expensive and unreliable as it covers relatively small groups of respondents. An alternative method implies the analysis of sales checks that are the digital footprints of a customer in a supermarket. The presented article suggests applying methods for hierarchical cluster analysis to distinguishing groups of complementary products. These methods imply the use of binary similarity measures. The article also presents the results of real-data experiments comparing the effectiveness of Jaccard index and chi-square measure and analyzes the methods for optimization of the number of clusters.