This project is conducted under the course Applied Forecasting in Complicated Systems - MSc Data Science @University of Amsterdam.
For large supermarkets like Walmart stores, forecasting future sales of products is crucial for keeping stock such that consumer demand can be met. This forecasting study focuses on the demand for a subcategory of hobby products in a Walmart store in California, USA. We implemented multiple models and compared their performance, including Prophet, Moving Average, Autoregressive Integrated Moving Average(ARIMA), Linear Regession, Gradient Boosting, Random Forest, Multi-Layer Perceptron (MLP) Regression, and a vanilla Long Short-Term Memory(LSTM) model.