Data mining techniques and machine learning model for Walmart weekly sales forecast

Abstract

The ability to forecast data accurately is extremely valuable in a vast array of domains such as health, sales, finance, weather or sports. Presented here is the study and implementation of data mining techniques and ensemble regression algorithm employed on sales data, consisting of weekly retail sales numbers from different departments in Walmart retail stores all over the United States of America over the period of 3 years with pre-holiday and holiday data presenting a spike in sales. The model implemented for prediction is Random. The metric to evaluate the model was the Mean Absolute Error (MAE) value. An analysis was performed to evaluate the model and its ability to forecast accurately. It is also notable that artificial neural networks can improve the performance and achieve highly accurate results. Key Terms ⎯ Machine Learning, Mean Absolute Error, Neural Networks, Random Forest, Sales Forecasting.