Demand Forecasting Using Artificial Neural Networks-A Case Study of American Retail Corporation

Artificial neural networks (ANNs) provide a way to make intelligent decisions while leveraging on today's processing power. In this paper, an attempt has been made to use ANN in demand forecasting by modeling it mathematically. MATLAB and R software are used to create the neural networks. Data has been organized and results are compared using Python. The complete analysis has been done using demand forecasting of American multinational retail corporation, Walmart. What we have managed to achieve in the end is almost perfect accuracy in forecasting demand of Walmart by ensuring that the set of inputs are complete enough to provide an output and then further ensuring that we do obtain an output. In compliance with the same, average sales of each Walmart store in question was calculated from training data and normalized. A correction factor was used to compensate for the effect of seasonality which is an external factor. By doing this, the model is saved from the trouble of having to map an extra factor which can otherwise be easily compensated for. The method used is a multi-layered perceptron in all cases. Iterations were done to find the best parameters to build the model.