Inventory Control Of Intermittent Demand Combined With Leading Economic Indicators

Intermittent demand is a discrete demand pattern with many zero demands between the non-zero demands. This demand pattern in some real production occasions(e.g., the demand for parts in heavy machinery industries) are correlated with some economic criteria(e.g., Construction Machinery Manufacturing spending). These features of the demand make it difficult to accurately forecast the demand and set up proper inventory policies.This paper introduces a forecasting and inventory control strategy for intermittent demand, which has strong correlation with leading economic indicators(LI). LI are discretized into different regimes. The demand is modeled as several(depending on how many regimes the demand is discretized) Zero-Inflated Poisson(ZIP) distributions with different parameters corresponding to different regimes. With historical data, the parameters of the forecasted distribution for the next period are generated using the Maximum Likelihood Estimation. And then the inventory position is decided using the method of dynamic programming according to the current state information, including the LI information. In this way, the historical demand information combined with LI can help to more accurately forecast the demand and thus reduce the cost of the inventory system.The improvement of using LI compared to the method without it depends on the correlation between the demand and the LI. With the increase in correlation, a higher improvement in forecasting accuracy and a larger reduction in the cost of the inventory system are obtained using LI. Moreover, this model achieves a lower cost compared with Graves(1997). And with the increase of the demand intermittency, a larger reduction in the cost of the inventory system is obtained. This model is especially useful for slow moving parts with intermittent demand pattern and high penalty cost, and could only be applied when the correlation coefficient is strong enough is always positive or negative over time.