Optimizing Production Schedules: A Hybrid Approach Using Neural Networks

Article & Journal: A new neuro-dominance rule for single-machine tardiness problem with double due date Published in: Neural Computing and Applications

Study Summary: In this study, we tackled the complex single-machine tardiness problem by integrating artificial intelligence with traditional dominance rules. We proposed a new "Neuro-Dominance Rule" (NDR) that utilizes an Artificial Neural Network (ANN) to predict and select the most effective dominance rules for varying scheduling conditions. This hybrid approach significantly reduced calculation times while maintaining high solution quality compared to classical methods.

Behind the Research: Scheduling problems are notoriously difficult (NP-hard) and traditional mathematical models often struggle with computation time as the problem size grows. We realized that static rules were not enough. By training a neural network to understand the characteristics of the job set, we created a dynamic system that "learns" how to schedule more efficiently. This work was an early step in demonstrating how machine learning could augment, rather than replace, fundamental operational research techniques.

Citation & DOI: Cakar, T., Köker, R., & Canay, O. (2015). A new neuro-dominance rule for single-machine tardiness problem with double due date. Neural Computing and Applications, 26(6), 1439-1450. DOI: 10.1007/s00521-014-1789-4