Bootstrapping to solve the limited data problem in production control: an application in batch process industries

Batch process industries are characterized by complex precedence relationships among operations, which makes the estimation of an acceptable workload very difficult. Previous research indicated that a regression-based model that uses aggregate job set characteristics may be used to support order acceptance decisions. Applications of such models in real-life assume that sufficient historical data on job sets and the corresponding makespans are available. In practice, however, historical data may be very limited and may not be sufficient to produce accurate regression estimates. This paper shows that such a lack of data significantly impacts the performance of regression-based order acceptance procedures. To resolve this problem, we devised a method that uses the bootstrap principle. A simulation study shows that performance improvements are obtained when using the parameters estimated from the bootstrapped data set, demonstrating that this bootstrapping procedure can indeed solve the limited data problem in production control.