Retail Sales Forecasting in E-Commerce Using Ensemble Learning

In this study, retail sales forecasting was conducted for Hepsiburada, one of Turkey’s leading e-commerce companies, using a ensemble learning approach. The dataset, prepared with feature engineering techniques, was tested with various methods such as Prophet, Exponential Smoothing, ARIMA, SARIMA, and Decision Trees. Daily sales forecasts were generated by combining the top three performing models. Compared to the traditional method, this approach demonstrated lower error rates and is expected to accelerate the process through automation while enabling more profitable future planning.