Pre-launch Insurance Customer Cluster Analysis using k-modes with Different Initialization Methods

  • Harish A S, Malathy C

Abstract

Insurance is a dynamic field where customer segmentation drives a key role in multiple business decisions such as customized product or premium pricing, gaining competitive advantage, marketing and targeting the right base for a balanced risk to profit ratio etc. Pre-launch customer data for insurance would mostly be higher in volume and categorical in nature with demographics or other behavioral data playing an important role in the outcome. This automatically makes k-modes clustering algorithm the primary choice, however the results of the algorithm could be affected by the choosing of various initialization methods or the dissimilarity calculation formulae used. In this paper, we apply and compare three different initialization methods for the centroids in k-modes algorithm before using it to compute the customer segments and study the impact of the differences in the outcome. This helps demonstrate the actual practical impact of using different centroid initialization methods and to understand if the results complement each other or affect the outcomes adversely.

Keywords: K-Modes, Insurance Analytics, Clustering, Cluster Initialization, Cluster Comparison

Published
2020-05-05
How to Cite
Harish A S, Malathy C. (2020). Pre-launch Insurance Customer Cluster Analysis using k-modes with Different Initialization Methods. International Journal of Advanced Science and Technology, 29(06), 2223 - 2232. Retrieved from https://sersc.org/journals/index.php/IJAST/article/view/13511