Makine Öğrenmesi Yöntemleri ile E-Ticarette Müşteri Kaybı Tahmini

Abstract

E-ticaret alanında işletmelerin varlığını sürdürebilme ve kâr edebilme durumları müşteri devamlılığının korunması ile gerçekleşmektedir. Yeni müşteri kazanma maliyetinin var olan müşterileri elde tutma maliyetinden daha yüksek olması, işletmenin yaşayabileceği olası müşteri kayıplarını tahmin edip buna yönelik önleyici stratejiler geliştirmesini bir öncelik haline getirmektedir. Bu çalışmada, makine öğrenmesi yöntemleri kullanılarak işletmelerin yaşayabileceği müşteri kayıp durumlarını tahmin etme, veriler üzerinde müşteri davranışlarını anlamayı sağlayan önemli özellikleri belirleme ve bu sonuçları iş stratejilerine dönüştürme hedeflenmektedir. Dört farklı sınıflandırma algoritması karşılaştırılarak e-ticaret veri setindeki müşteri davranışları incelenmiş ve tahminleme yapılmıştır. Hiper parametre optimizasyonu işlemleri Optuna ile gerçekleştirilmiş olup özellik seçim işlemlerinde SelectBest yöntemi ve istatiksel testler birlikte uygulanmıştır. Veri sızıntısını önlemek için gerekli protokoller kullanılmış ve model kararlılığı rastgele örnekleme yöntemiyle test edilmiştir. Performans değerlendirmesinde sonuçlar, Friedman ve Wilcoxon testleri ile doğrulanmış ve metriklerin % 95 güven aralıkları hesaplanmıştır. Model açıklanabilirliği için SHAP ve LIME yaklaşımları bir arada ele alınmıştır. Çalışma sonuçlarında modeller karşılaştırıldığında en dengeli performansı XGBoost algoritması sergilemiş olup, ona en yakın performansı Random Forest algoritması göstermiştir. Yapılan Friedman testi ile modeller arasında anlamlı farklılıklar ortaya konulmuş, Wilcoxon ikili teslerle ise XGBoost ve Random Forest’in diğer modellere üstünlüğü doğrulanmıştır. Müşteri kayıp durumlarını tahminlemede sonuçlar üzerindeki en belirleyici özelliklerin; üyelik süresi, cashback ödemeleri, en son ne zaman alışveriş yapıldığı ve şikayet durumları olmak üzere dört özellik olduğu tespit edilmiştir. Veri sızıntısı korumalı, Optuna tabanlı ve açıklanabilir makine öğrenmesi yaklaşımları; e-ticaret firmalarının risk segmentasyonu yapma, erken müdahale etme ve müşterilerin elde tutulması için stratejiler geliştirmesini desteklemektedir.

Keywords

• E-Ticaret
• Müşteri Kaybı Tahmini
• Makine Öğrenmesi
• Model Seçimi
• Performans Değerlendirmesi
• Model Açıklanabilirliği

Supporting Institution

Üsküdar Üniversitesi, İstanbul Topkapı Üniversitesi

Ethical Statement

Hazırlanan makalede etik kurul izni alınmasına gerek yoktur. Yazarlar, hazırlanan makalede herhangi bir kişi/kurum ile çıkar çatışması bulunmadığını beyan etmektedir.

Customer Churn Prediction in E-Commerce Using Machine Learning Methods

Abstract

In the e-commerce sector, businesses can only survive and remain profitable by retaining their customers. The fact that the cost of acquiring new customers is higher than the cost of retaining existing customers makes it a priority for businesses to predict potential customer losses and develop preventive strategies. This study aims to predict customer loss scenarios that businesses may experience using machine learning methods, identify key features that enable understanding customer behavior in the data, and translate these results into business strategies. Four different classification algorithms were compared to examine and predict customer behaviors in the e-commerce dataset. Hyperparameter optimization was performed using Optuna, and the SelectBest method and statistical tests were applied together in feature selection. Necessary protocols were used to prevent data leakage, and model stability was tested using random sampling. The results of the performance evaluation were validated using Friedman and Wilcoxon tests, and the 95% confidence intervals of the metrics were calculated. SHAP and LIME approaches were considered together for model interpretability. When comparing the models in the study results, the XGBoost algorithm demonstrated the most balanced performance, followed closely by the Random Forest algorithm. The Friedman test revealed significant differences between the models, while the Wilcoxon paired tests confirmed the superiority of XGBoost and Random Forest over the other models. It was determined that the most decisive features affecting the results in predicting customer churn were customer membership duration, cashback payments, the last time a purchase was made, and complaint status. Data leak-protected, Optuna-based, and explainable machine learning approaches support e-commerce companies in developing strategies for risk segmentation, early intervention, and customer retention.

Keywords

• E-Commerce
• Customer Churn Prediction
• Machine Learning
• Model Selection
• Performance Evaluation
• Model Interpretability

Supporting Institution

Uskudar University, Istanbul Topkapı University

Ethical Statement

No ethics committee approval is required for the prepared article. The authors declare that there is no conflict of interest with any person/institution in the prepared article.

References

1. [1] C. Wei and I. Chiu, “Turning telecommunications call details to churn prediction: a data mining approach,” Expert Systems with Applications, vol. 23, no. 2, pp. 103–112, 2002, doi:10.1016/S0957-4174(02)00030-1.
2. [2] G. W. Bock and M. Uncles, “A taxonomy of differences between consumers for segmentation purposes,” Journal of Marketing Management, 2002, doi:10.1016/S0167-8116(02)00081-2.
3. [3] K. Buckinx and D. Van den Poel, “Customer base analysis: partial defection of behaviourally loyal clients in a non-contractual FMCG retail setting,” European Journal of Operational Research, vol. 164, 2005, doi: 10.1016/j.ejor.2003.12.010.
4. [4] E. Ngai, L. Xiu, and D. Chau, “Application of data mining techniques in customer relationship management: A literature review and classification,” Expert Systems with Applications, vol. 36, 2009, doi:10.1016/j.eswa.2008.02.021
5. [5] J. Sun, H. Li, H. Fujita, B. Fu, and W. Ai, “Class-imbalanced dynamic financial distress prediction based on Adaboost-SVM ensemble combined with SMOTE and time weighting,” Information Fusion, vol. 54, pp. 128–144, 2020, doi: 10.1016/j.inffus.2019.07.006.
6. [6] T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning, Springer, 2009, doi:10.1007/978-0-387-84858-7.
7. [7] J. Kotsiantis, “Supervised machine learning: A review of classification techniques,” Informatica, vol. 31, 2007. [Online]. Available: https://www.researchgate.net/publication/220166738_Supervised_Machine_Learning_A_Review_of_Classification_Techniques.
8. [8] H. He and E. Garcia, “Learning from imbalanced data,” IEEE Transactions on Knowledge and Data Engineering, vol. 21, no. 9, pp. 1263–1284, 2009, doi:10.1109/TKDE.2008.239.

9. [9] S. Lundberg and S. Lee, “A unified approach to interpreting model predictions,” Advances in Neural Information Processing Systems, 2017, doi:10.48550/arXiv.1705.07874.
10. [10] M. Ribeiro, S. Singh, and C. Guestrin, “Why should I trust you? Explaining the predictions of any classifier,” Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, doi:10.1145/2939672.2939778.
11. [11] F. Kayaalp and M. S. Başarslan, “Shiny ile R uygulaması geliştirme ve müşteri kayıp analizine ilişkin web destekli bir örnek çalışma,” Duzce University Journal of Science and Technology, vol. 5, no. 2, pp. 538–547, 2017. [Online]. Available: https://dergipark.org.tr/en/pub/dubited/article/313803.
12. [12] Y. Balcıoğlu and B. Sezen, “Yapay sinir ağları ile banka müşterilerinin bankadan ayrılma olasılığının tahmini,” Business and Organization Research, Izmir, 2019. [Online]. Available: https://www.researchgate.net/publication/335727610_YAPAY_SINIR_AGLARI_ILE_BANKA_MUSTERILERININ_BANKADAN_AYRILMA_OLASILIGININ_TAHMINI.
13. [13] A. B. Zoric, “Predicting customer churn in banking industry using neural networks,” Interdisciplinary Description of Complex Systems, vol. 14, no. 2, pp. 116–124, 2016.
14. [14] A. J. A. Shahmi, “Customer churn prediction in e-commerce using ensemble learning methods,” 2025. [Online]. Available: https://www.researchgate.net/publication/389320064.
15. [15] H. T. Kazim, “Machine learning approaches for e-commerce churn analysis,” 2025. [Online]. Available: https://repository.rit.edu/theses/12162/
16. [16] H. Ren, “Machine learning-based prediction of customer churn risk in e-commerce,” 2025, doi:10.54254/2754-1169/2024.19473.
17. [17] S. Kumar, S. Deep, and P. Kalra, “A comprehensive analysis of machine learning techniques for churn prediction in e-commerce: A comparative study,” International Journal of Computer Trends and Technology, vol. 72, no. 5, pp. 163–170, 2024. doi: 10.14445/22312803/IJCTT-V72I5P119.
18. [18] Y. Sarıyıldız, “E-ticaret ve dijital pazarlama sektöründe makine öğrenmesi algoritmaları kullanılarak müşteri kaybı tahmini,” 2024. [Online]. Available: https://www.proquest.com/docview/3250849129.
19. [19] H. Uzel, “An explainable hybrid deep learning approach for power quality disturbance classification,” Bulletin of the Polish Academy of Sciences: Technical Sciences, vol. 74, no. 2, e157326, 2026, doi: 10.24425/bpasts.2026.157326.
20. [20] F. Alpsalaz, Y. Özüpak, E. Aslan, and H. Uzel, “Hybrid machine learning approach for enhanced fault detection and power estimation in photovoltaic systems,” IET Renewable Power Generation, vol. 20, no. 1, 2026, doi: 10.1049/rpg2.70153.
21. [21] E. Aslan and Y. Özüpak, “Fault Detection in Electricity Transmission Lines: A Comparative Analysis of Ensemble Machine Learning and Convolutional Neural Network Methods,” Firat University Journal of Experimental and Computational Engineering, vol. 5, no. 1, pp. 150–168, 2026, doi: 10.62520/fujece.1581543
22. [22] A. VERMA, “Ecommerce Customer Churn Analysis and Prediction,” Kaggle Dataset, 2021. [Online]. Available: Kaggle dataset path: ankitverma2010/ecommerce-customer-churn-analysis-and-prediction. Accessed: 06-Ocak-2026.
23. [23] F. E. Harrell Jr., Regression Modeling Strategies: With Applications to Linear Models, Logistic and Ordinal Regression, and Survival Analysis, 2nd ed. Cham, Switzerland: Springer, 2015, doi:10.1007/978-3-319-19425-7.
24. [24] H. B. Mann and D. R. Whitney, “On a test of whether one of two random variables is stochastically larger than the other,” Annals of Mathematical Statistics, vol. 18, pp. 50–60, 1947, doi: 10.1214/aoms/1177730491.
25. [25] K. Pearson, “On the criterion that a given system of deviations from the probable in the case of a correlated system of variables is such that it can be reasonably supposed to have arisen from random sampling,” Philosophical Magazine, ser. 5, vol. 50, pp. 157–175, 1900, doi: 10.1080/14786440009463897.
26. [26] Y. Benjamini and Y. Hochberg, “Controlling the false discovery rate: A practical and powerful approach to multiple testing,” Journal of the Royal Statistical Society: Series B (Methodological), vol. 57, pp. 289–300, 1995, doi: 10.1111/j.2517-6161.1995.tb02031.x.
27. [27] F. Pedregosa et al., “Scikit-learn: Machine learning in Python,” Journal of Machine Learning Research, vol. 12, pp. 2825–2830, 2011. [Online]. Available: https://doi.org/10.48550/arXiv.1201.0490.
28. [28] T. M. Cover and P. E. Hart, “Nearest neighbor pattern classification,” IEEE Transactions on Information Theory, vol. 13, no. 1, pp. 21–27, 1967, doi: 10.1109/TIT.1967.1053964.
29. [29] L. Breiman, “Bagging predictors,” Machine Learning, vol. 24, no. 2, pp. 123–140, 1996, doi: 10.1007/BF00058655.
30. [30] L. Breiman, “Random forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, 2001, doi: 10.1023/A:1010933404324.
31. [31] T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” in Proc. 22nd ACM SIGKDD Int. Conf. Knowledge Discovery and Data Mining (KDD), San Francisco, CA, USA, Aug. 13–17, 2016, pp. 785–794, doi: 10.1145/2939672.2939785.
32. [32] J. Snoek, H. Larochelle, and R. P. Adams, “Practical Bayesian optimization of machine learning algorithms,” in Advances in Neural Information Processing Systems (NIPS), 2012, pp. 2960–2968.
33. [33] M. Emeç and M. H. Özcanhan, “Makine öğrenmesi algoritmalarında hiper parametre belirleme,” in Mühendislikte Öncü ve Çağdaş Çalışmalar, pp. 71–98, 2023. [Online]. Available: https://www.researchgate.net/publication/372496761.
34. [34] J. Bergstra, R. Bardenet, Y. Bengio, and B. Kégl, “Algorithms for hyper-parameter optimization,” in Proc. 24th Int. Conf. Neural Information Processing Systems (NIPS), Granada, Spain, Dec. 12–15, 2011, pp. 2546–2554. [Online]. Available: https://proceedings.neurips.cc/paper_files/paper/2011/file/86e8f7ab32cfd12577bc2619bc635690-Paper.pdf
35. [35] V. Sinap, “Optuna tabanlı hiper parametre optimizasyonu ile konut fiyat tahminlemede makine öğrenmesi tekniklerinin karşılaştırmalı analizi,” Gazi Üniversitesi Fen Bilimleri Dergisi Part C: Tasarım ve Teknoloji, vol. 13, no. 1, pp. 10–28, 2025, doi:10.29109/gujsc.1544987.
36. [36] Student, “The probable error of a mean,” Biometrika, vol. 6, no. 1, pp. 1–25, 1908, doi: 10.2307/2331554.
37. [37] M. Friedman, “The use of ranks to avoid the assumption of normality implicit in the analysis of variance,” Journal of the American Statistical Association, vol. 32, no. 200, pp. 675–701, 1937, doi: 10.1080/01621459.1937.10503522.
38. [38] S. Holm, “A simple sequentially rejective multiple test procedure,” Scandinavian Journal of Statistics, vol. 6, no. 2, pp. 65–70, 1979. [Online]. Available: http://www.jstor.org/stable/4615733.
39. [39] A. Alshamsi, “Customer churn prediction in ecommerce sector,” Master’s thesis, Rochester Institute of Technology, Rochester, NY, USA, 2022. [Online]. Available: https://repository.rit.edu/theses/11183/
40. [40] M. A. Al Rahib, N. Saha, R. Mia, and A. Sattar, “Customer data prediction and analysis in e-commerce using machine learning,” Bulletin of Electrical Engineering and Informatics, vol. 13, no. 4, pp. 2624–2633, 2024. doi: 10.11591/eei.v13i4.6420.
41. [41] B. Yilmaz, “Application of customer’s churn prediction models in e-commerce,” Bachelor’s thesis, Vrije Universiteit Amsterdam, Amsterdam, Netherlands, 2025. [Online]. Available: https://www.cs.vu.nl/~wanf/theses/yilmaz-bscthesis.pdf