Optimizing Clustering Models Using Principle Component Analysis for Car Customers

Abstract

 In the competitive business world, companies strategically utilize customer data to achieve goals, requiring a comprehensive understanding of various customer traits, behaviors and needs. Customer segmentation, an important strategy, requires grouping individuals based on various characteristics. The K-Means algorithm is widely used for customer data grouping connectivity because of its ease of implementation in Machine Learning. However, challenges arise in high-dimensional data, prompting the need for dimensionality reduction. Principal Component Analysis (PCA) is emerging as an effective method for data communication while minimizing information loss. Previous research emphasizes the success of PCA in improving analysis and clustering efficiency. This research contributes by integrating PCA into K-Means clustering to analyze customer segments in a car company. This empowers companies to attract new customers, implement targeted marketing, understand customer-company relationships, and increase expected profitability. PCA, which preserves 75% of the variation with 3 principal components, precedes the implementation of K-Means after normalization. Evaluation using the Elbow and Silhouette Score Method identified eight optimal clusters. The post-PCA K-Means model with optimal cluster selection produces a Silhouette Score of 0.7789.