Integration of Pre-processing and Machine Learning Methods for Improved Customer Segmentation

Abstract

In a highly competitive business environment, the success of companies is largely dependent on the effective analysis of customer data. This study investigates how machine learning algorithms can be applied to customer segmentation and explores the potential advantages this application can offer businesses. In our research, three different clustering algorithms, including K-means, hierarchical clustering, and spectral clustering, were utilized to perform segmentation analysis on a customer data set containing 27 attributes. Various data pre-processing techniques, including dimension reduction, were employed throughout the analysis process to enhance the segmentation's accuracy and improve the model's efficacy. The results have been supported by various visualization techniques, enabling detailed reporting that can be utilized in decision-making processes. The tests indicate that the spectral clustering algorithm achieved the highest clustering accuracy with Silhouette score of approximately 0.61. Additionally, an improvement of about 0.29 was reported when compared to the performance before pre-processing. © 2024 IEEE.