Implementation of models for Demand forecasting for e-commerce using time series forecasting

The study delves into the significance of demand forecasting within the e-commerce realm, crucial for businesses to anticipate future customer needs and optimize inventory management and resource allocation strategies. Focusing on employing time series forecasting methodologies, specifically XGBOOST and SARIMAX (Seasonal AutoRegressive Integrated Moving Average with eXogenous factors) models, the research aims to predict demand accurately in e-commerce scenarios. By analyzing historical sales data and integrating pertinent external variables like promotional events and seasonal trends, these models endeavor to forecast forthcoming demand trends for a variety of products or services offered by e-commerce platforms. The investigation scrutinizes the intricacies of XGBOOST and SARIMAX models, highlighting their merits and limitations in the context of e-commerce demand forecasting. Moreover, the study assesses model performance using metrics like Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) to gauge accuracy and reliability in demand prediction. The insights gleaned from this study contribute to advancing demand forecasting techniques in e-commerce, offering valuable guidance for businesses aiming to refine operational efficiency and elevate customer satisfaction through precise demand projections. The SARIMA model deployed in this work has a MAPE value of 0.027 which is better than any previously formed SARIMA models on this particular dataset. The XGBOOST model implemented in this work has RMSE value of 2852.098 which is lesser than any of the work done on this dataset before. Also the calculated accuracy comes out to be 98.348% for XGBOOST which is remarkable and is best on this data.