Intermittent demand forecast modeling

Forecasting helps us by casting light on our projected path between today and the future. Efficient forecasting leads to a highly productive environment. Demand forecasting involves selecting a proper foresting methodology for each product. Our research focuses on selecting proper forecasting models by using historical demand data from a confectionary producer. The data is collected from sales vouchers which list product type, customer id, quantity and data for each sale transaction. The data is highly irregular, intermittent in nature. Preselected set of forecasting models are used to make demand projections. The model set includes simple moving averages, exponential smoothing, and Croston's method. The performance of each model is determined based on an error metric which measures the deviation between projected and actual values. This study is intended to present a new and alternative method that could be used in production planning.