Eyeballing heuristics for dynamic lot sizing problems with rolling horizons

Abstract

The dynamic lot sizing problem with rolling horizons was found to be a difficult problem for getting efficient solutions. The existing algorithms do not perform well for short forecast windows because of uncertain demands beyond the forecast windows. The one-way eyeballing heuristics is introduced to solve the problem. It determines the production cycle if the eyeballing comparison ends before a forecast window, otherwise, it generates a safety stock to satisfy some demands after the forecast window. The computational results on uniform and normal demand patterns show a big improvement in costs for problems with short forecast windows. We have also applied the heuristics to dynamic lot sizing problems with variable forecast windows and also obtained good computational results.