Application of Clustering and Classification Algorithms in Analyzing Customer Behavior in Data-Driven Marketing: A Case Study of Amazon Customers

In data-driven marketing, customer behavior analysis plays a crucial role in developing targeted marketing strategies aimed at increasing return on investment, enhancing profitability, and gaining a larger market share. In this study, four clustering methods- including K-means, density-based clustering, principal component analysis, and hierarchical clustering- as well as four classification methods- including Support Vector Machine, XGBoost, Random Forest, and Gradient Boosting- are examined for customer behavior analysis. The data for this study was extracted from the "Amazon Customer Behavior Survey" dataset, which includes 23 features from 602 customers. Initially, the data was preprocessed, and then, using clustering methods, customers were divided into different groups. The performance of these methods was evaluated based on criteria such as the silhouette index, and ultimately, appropriate marketing strategies for each cluster were proposed. Additionally, to examine the possibility of predicting customer membership in the extracted clusters, the aforementioned classification models were implemented and compared. The results indicate that the K-means method performed the best in clustering, while the XGBoost model performed the best in classification. The innovation of this research lies in combining clustering and classification methods to provide targeted marketing strategies and comprehensively comparing these methods on real customer data. This study demonstrates that combining clustering and classification methods can help businesses better understand customer behavior and make more optimal marketing decisions.