By combining ARIMA models with XGBoost in a single model ensemble we manage to achieve better model performance.
The dataset contains historical product demand for a manufacturing company with footprints globally. The company provides thousands of products within dozens of product categories. There are four central warehouses to ship products within the region it is responsible for. Since the products are manufactured in different locations all over the world, it normally takes more than one month to ship products via ocean to different central warehouses. If forecasts for each product in different central with reasonable accuracy for the monthly demand for month after next can be achieved, it would be beneficial to the company in multiple ways.
The dataset was pulled from Kaggle, this is the link: https://www.kaggle.com/felixzhao/productdemandforecasting