<![CDATA[The rapid development of the internet is one of the driving factors behind the growth of e-commerce. This has led to the emergence of many e-commerce companies, resulting in intense competition among them. Customers have the right to choose the e-commerce platforms that suit their needs and can switch to competing e-commerce platforms, a phenomenon known as customer churn. This issue can be addressed by classifying customer behavior based on existing data. This study utilizes the Random Forest Classifier method, employing the SMOTE and SMOTEENN resampling techniques to handle data imbalance. From the conducted research, the best results were achieved using the SMOTE implementation, with an accuracy of 96.3%, precision of 87.8%, recall of 87.1%, f1-score of 87.4%, and an AUC score of 93%. These results successfully strike a balance between recognizing the positive class (churn) and controlling false positives. On the other hand, the SMOTEENN implementation yields the best recall value and an increase in AUC score, but it comes with a significant decrease in precision, indicating a challenge in controlling false positives.]]>
Copyrights © 2023
