John Boylan and I have been working on a paper about state-space models for intermittent data. We have had some good progress in that direction and have submitted the paper to IJF. Although it is still under review, we decided to publish the working paper in order to promote the thing. Here’s the abstract:
Intermittent demand forecasting is an important supply chain task, which is commonly done using methods based on exponential smoothing. These methods however do not have underlying statistical models, which limits their generalisation. In this paper we propose a general state-space model that takes intermittence of data into account, extending the taxonomy of exponential smoothing models. We show that this model has a connection with conventional non-intermittent state space models and underlies Croston’s and Teunter-Syntetos-Babai (TSB) forecasting methods. We discuss properties of the proposed models and show how a selection can be made between them in the proposed framework. We then conduct experiments on simulated data and on two real life datasets, demonstrating advantages of the proposed approach.
By the way, all the models discussed in the paper, are available in the smooth package for R. All the functions in the package allow switching between Croston style, TSB style probabilities or even select the most appropriate one based on information criteria. I will make a post on that thing in a month or so.