Minimizing the error bound for the dynamic lot size model

In the dynamic lot size model, the production plan for the first few periods is usually determined based on the forecast data for some fixed data horizon. This fixed data horizon may not be long enough, in the sense that information beyond that data horizon may affect the optimal decisions in the first few periods. In the literature there are two approaches to deal with this insufficiency of information: to extend the data horizon or to find a worst-case bound on the error induced by imposing a finite data horizon on the model. This paper takes the second approach. When the second approach is taken, existing papers only evaluate the error bound after the production plan of the first few periods is determined. By contrast, this paper determines the production plan such that the corresponding error bound is minimal. The authors provide a polynomial time algorithm for the problem under the realistic assumption that a ‘speculative motive’ cost structure is not allowed.