Temporal Customer Segmentation Using the Self-organizing Time Map

Abstract

Visual clustering provides effective tools for understanding relationships among clusters in a data space. This paper applies an adaptation of the standard Self-Organizing Map for visual temporal clustering in exploring the customer base and tracking customer behavior of a department store over a 22-week period. In contrast to traditional clustering techniques, which often provide a static snapshot of the customer base and overlook the possible dynamics, the Self-Organizing Time Map enables exploring complex patterns over time by visualizing the results in a user-friendly way. We demonstrate the effectiveness of the application using department store data with more than half a million rows of weekly aggregated customer information.