<![CDATA[E-commerce businesses face the challenge of retaining customers in the era of rapid digital expansion. Customer churn prediction becomes essential for strategic decision-making by offering insights into potential revenue loss and customer loyalty. One of the problem in customer churn prediction comes from the presence of outliers in the data. This research delves into seeing the effects on churn prediction f1-score by incorporating a combination of techniques including outlier detection via k-means clustering and DBSCAN, as well as employing XGBoost and Catboost as classifiers. Results indicate that using Catboost gives a better performance of 96% F1-Score for e-commerce customer churn dataset with outliers, and removing outliers does not result in an increase in performance]]>
Copyrights © 2024
