A Hybrid Demand Forecasting for Intermittent Demand Patterns using Machine Learning Techniques

Abstract

Intermittent demand forecasting is a major obstacle to the digital supply chain revolution and accurate demand forecasting (DF) enhances the overall productivity of enterprises and minimizes expenditures. Random and low-volume demand is referred to as intermittent demand. It appears erratically between demand periods, with a high percentage of zero values. Companies managing complex inventory systems face difficulties due to intermittent demand's unexpected nature, which results in expenses associated with excess inventory or stockouts. This research aims to present an integrated forecasting strategy for intermittent demand problems using Long Short-Term Memory (LSTM) and parametric methods to assist new range of management decisions as a vital element of intelligent supply chain. This can be achieved by two biggest e- commerce companies' dataset and the results showed that hybrid forecasting framework results are more accurately than traditional forecasting techniques. The hybrid Intermittent forecasting can be used as an alternate technique for anticipating intermittent demand since its ease of computing of forecasting results accurately than the classical forecasting methods.