Evaluation of Clustering with PCA for Market Segmentation: A Study Using Simulated and Surrogate Data

Market segmentation is a strategy crucial for manufacturing enterprises, enabling targeted approaches and actions aligned with specific business objectives. By segmenting the market, companies can identify profitable segments and gain a competitive advantage through superior market positioning. Principal Component Analysis (PCA) is a common statistical technique used for dimensionality reduction and feature extraction, and PCA-based clustering is a data-driven approach applied for identifying market segments. However, effective segmentation requires systematic application of methods, appropriate selection of principal components, and proper clustering validation measures. In this paper, we propose a framework to evaluate PCA-based clustering for segmentation using simulated and surrogate data along with several internal and external validation clustering measures. This framework can be applied to dimensionality reduction methods to obtain stable clustering solutions, which can be effectively optimized using various validation measures; it can also provide explainable insight into segments identified by clustering with PCA. To illustrate the effectiveness of our proposed framework, we present a case study. We simulate clustering data and use e-car data as surrogate data to evaluate the clustering results, employing k-means as the clustering method.