Development of Machine Learning-Based Sales Cancellation/Return Forecasting Models for the E-Commerce Industry

E-commerce is evolving rapidly, creating a more competitive market environment. With this development, gaining a competitive advantage has become even more crucial. Companies implement various strategic moves to maintain their position in this competitive market. Strategies and forecasts have a high ranking among these strategies. Predicting sales cancellations and returns is crucial for companies to anticipate future challenges and stay ahead. The aim of this study is to develop sales cancellation/return prediction models for the e-commerce sector. To achieve this, sales cancellation/return prediction models have been generated using Multi-Layer Perceptron (MLP), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), and Logistic Regression (LR). A weekly dataset has been created using 4909 rows of sales cancellation/return data. The performance of the models has been evaluated using precision, recall, F1 score, and accuracy. Among all the methods, it has been observed that RF and XGBoost delivered the best performance.