Machine Learning and Inverse Optimization for Estimation of Weighting Factors in Multi-Objective Production Scheduling Problems
Abstract
:1. Introduction
2. Literature Review
- A priori methods: The decision maker specifies the preferred objective function before executing the solution process.
- Interactive methods: Phases of interaction between decision-makers and the solution process are iteratively conducted.
- Posteriori or generation methods: After the solutions are generated by the weighting method or -constraint method, the decision-maker selects the preferred solution.
3. Problem Description
3.1. Definition of Multi-Objective Parallel Machine Scheduling Problem
- There is no idle time set for each machine.
- One machine can only handle one job at a time.
- Each job cannot be interrupted during processing. No preemption is allowed.
3.2. Weighting Factor Estimation Problem from Historical Data
- When the plant is operated by human experts, they will set appropriate weighting factors in daily scheduling. The derived weighting factors can be used to understand the expert knowledge of human operators.
- The solutions of the scheduling system can be used to set appropriate weighting factors of the objective function when an automated scheduling system is equipped in a real factory.
4. Weighting Factor Estimation Method from Scheduling Results
4.1. Machine Learning
4.2. Feature Extraction
4.2.1. Small Scale Problems with 2 Objectives Optimization Problems
4.2.2. Large Scale Problems with 2 Objectives Optimization Problems
4.2.3. Small Scale Problems with 3 Objectives Optimization Problems
- maximum completion time + sum of weighted delivery delay + sum of setup costs
- sum of weighted completion time + maximum weighted delivery delay + sum of setup costs
4.3. Feature Selection
- Start with an empty set and add the feature that improves accuracy the most to the subset .
- Add the feature that improves accuracy the most when adding features to subset . The MSE is computed when the feature is added.
- If the state of Equation (11) is maintained, go to 2. Otherwise, exit.
4.4. Inverse Optimization
Algorithm 1: Inverse optimization |
5. Computational Experiments
5.1. Effectiveness of Features in Machine Learning
5.2. Comparison of Each Case in Machine Learning
5.3. Comparison of Each Case in Inverse Optimization
5.4. Comparison of Machine Learning Method and Inverse Optimization Method
5.5. Application to Chemical Batch Plants
5.6. Discussion
- In randomly generated problems, the estimation accuracy of the large-scale problem was higher than that of the small-scale problem because the solution changed continuously in response to the change of the weighting factor.
- In the case of machine learning, the feature of the variance of completion time for each machine was effective for small-scale problems, and the features of the sum of completion time of each machine and the variance of delivery time setting for each machine were effective for large-scale problems.
- When using the input/output data of the same scheduling model, if a solution close to the exact solution is used, the inverse optimization has higher estimation accuracy than machine learning because the gradient was recalculated many times during the estimation process.
6. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
Abbreviations
AHP | analytic hierarchy process |
MSE | mean squared error |
SFS | sequential forward selection |
SBS | sequential backward selection |
Symbol | |
MSE | |
Number of data | |
Estimated weighting factor | |
True weighting factor | |
Feature extraction | |
Number of jobs | |
Number of machines | |
Weight of objective function f | |
Weight of job i | |
Processing time at job at machine | |
Setup cost when switching from job i to job j | |
Completion time of job i at machine k | |
Delay in delivery of job i at machine k | |
1 when job j is processed immediately after job i at machine k, 0 otherwise | |
1 when job i is processed at machine k, 0 otherwise | |
Maximum completion time of machine | |
Mean of completion time of each machine | |
Total delivery date of job processed of machine | |
Mean of the delivery date of job processed by each machine | |
Starting time order of job in ascending order | |
Weight order of job in ascending order | |
Processing time order of job in ascending order | |
Delivery time order of job in ascending order | |
order of job in ascending order | |
order of job in ascending order | |
order of job in ascending order | |
order of job in ascending order | |
Variation of the completion time | |
Spearman’s rank correlation coefficients | |
Variance of delivery time setting for each machine | |
Sum of completion time of each machine | |
Shortest processing time selectivity | |
Feature selection | |
Feature | |
Subset of the selected features | |
Feature that improves accuracy the most when adding features to subset | |
Subset of the features when feature is added | |
Constant expressing the degree of tolerance | |
Number of selected features | |
Inverse optimization | |
Initial weighting factor | |
Maximum number of iterations | |
Solutions with known weighting factor | |
Learning rate | |
Number of problem instances |
References
- Provost, F.; Fawcett, T. Data Science and its Relationship to Big Data and Data-Driven Decision Making. Big Data 2013, 1, 51–59. [Google Scholar] [CrossRef] [PubMed]
- Kamble, S.S.; Gunasekaran, A.; Gawankar, S.A. Achieving sustainable performance in a data-driven agriculture supply chain: A review for research and applications. Int. J. Prod. Econ. 2020, 219, 179–194. [Google Scholar] [CrossRef]
- Zhang, J.; Wang, K. Data-Driven Intelligent Transportation Systems: A Survey. IEEE Trans. Intell. Transp. Syst. 2011, 12, 1624–1639. [Google Scholar] [CrossRef]
- Whitt, W.; Zhang, X. A data-driven model of an emergency department. Oper. Res. Health Care 2017, 12, 1–15. [Google Scholar] [CrossRef]
- Vaccari, M.; di Capaci, R.B.; Brunazzi, E.; Tognotti, L.; Pierno, P.; Vagheggi, R.; Pannocchia, G. Implementation of an Industry 4.0 system to optimally manage chemical plant operation. IFAC-PapersOnLine 2020, 53, 11545–11550. [Google Scholar] [CrossRef]
- Vaccari, M.; di Capaci, R.B.; Brunazzi, E.; Tognotti, L.; Pierno, P.; Vagheggi, R.; Pannocchia, G. Optimally Managing Chemical Plant Operations: An Example Oriented by Industry 4.0 Paradigms. Ind. Eng. Chem. Res. 2021, 60, 7853–7867. [Google Scholar] [CrossRef]
- Burnak, B.; Pistikopoulos, E.N. Integrated process design, scheduling, and model predictive control of batch processes with closed-loop implementation. AIChE J. 2020, 66, e16981. [Google Scholar] [CrossRef]
- Nishi, T.; Sakata, A.; Hasebe, S.; Hashimoto, I. Autonomous decentralized scheduling system for just-in-time production. Comput. Chem. Eng. 2000, 24, 345–351. [Google Scholar] [CrossRef]
- Nishi, T.; Konishi, M.; Hasebe, S. An autonomous decentralized supply chain planning system for multi-stage production processes. J. Intell. Manuf. 2005, 16, 259–275. [Google Scholar] [CrossRef]
- Zhou, G.; Jiang, P.; Huang, G.Q. A game-theory approach for job scheduling in networked manufacturing. Int. J. Adv. Manuf. Technol. 2009, 41, 972–985. [Google Scholar] [CrossRef]
- Lee, J.W.; Kim, M.K. An Evolutionary Game theory-based optimal scheduling strategy for multiagent distribution network operation considering voltage management. IEEE Access 2022, 10, 50227–50241. [Google Scholar] [CrossRef]
- Taehyeun, P.; Walid, S. Kolkata Paise Restaurant Game for resource allocation in the Internet of Things. In Proceedings of the 2017 51st Asilomar Conference on Signals, Systems, and Computers, Pacific Grove, CA, USA, 29 October–1 November 2017; pp. 1774–1778. [Google Scholar] [CrossRef]
- Venkatachalam, A.; Christos, T.M. State estimation in online batch production scheduling: Concepts, definitions, algorithms and optimization models. Comput. Chem. Eng. 2021, 146, 107209. [Google Scholar] [CrossRef]
- Baker, K.R.; Smith, J.C. A Multiple-Criterion Model for Machine Scheduling. J. Sched. 2003, 6, 7–16. [Google Scholar] [CrossRef]
- Aoki, Y. Multi-Objective Optimization Method. Oper. Res. 1978, 23, 511–516. [Google Scholar]
- Georgiadis, G.R.; Elekidis, A.R.; Georgiadis, M.C. Optimization-based scheduling for the process industries: From theory to real-life industrial applications. Processes 2019, 7, 438. [Google Scholar] [CrossRef]
- Watanabe, S.; Hiroyasu, T.; Miki, M. Neighborhood Cultivation Genetic Algorithm for Multi-Objective Optimization Problems. J. Inf. Process. 2002, 43, 183–198. [Google Scholar]
- Kobayashi, H.; Tachi, R.; Tamaki, H. Introduction of Weighting Factor Setting Technique for Multi-Objective Optimization. J. Inst. Syst. Control. Inf. Eng. 2018, 31, 281–294. [Google Scholar] [CrossRef]
- Ghobadi, K.; Lee, T.; Mahmoudzadeh, H.; Terekhov, D. Robust inverse optimization. Oper. Res. Lett. 2018, 46, 339–344. [Google Scholar] [CrossRef]
- Zou, Q.; Zhang, Q.; Yang, J.; Gragg, J. An inverse optimization approach for determining weights of joint displacement objective function for upper body kinematic posture prediction. Robotica 2012, 30, 389–404. [Google Scholar] [CrossRef]
- Thomas, L.S. A Scaling Method for Priorities in Hierarchical Structures. J. Math. Psychol. 1997, 15, 234–281. [Google Scholar] [CrossRef]
- Ic, Y.T.; Yurdakul, M.; Eraslan, E. Development of a Component-Based Machining Centre Selection Model using AHP. Int. J. Prod. Res. 2012, 50, 6489–6498. [Google Scholar] [CrossRef]
- Cortés, B.M.; García, J.C.E.; Hernández, F.R. Multi-Objective Flow-Shop Scheduling with Parallel Machines. Int. J. Prod. Res. 2012, 50, 2796–2808. [Google Scholar] [CrossRef]
- Matsuoka, Y.; Nishi, T.; Tierney, K. Machine Learning Approach for Identification of Objective Function in Production Scheduling Problems. In Proceedings of the 2019 IEEE 15th International Conference on Automation Science and Engineering (CASE), Vancouver, BC, Canada, 22–26 August 2019; pp. 679–684. [Google Scholar] [CrossRef]
- Wu, M.; Kwong, S.; Jia, Y.; Li, K.; Zhang, Q. Adaptive Weights Generation for Decomposition-Based Multi-Objective Optimization Using Gaussian Process Regression. In Proceedings of the GECCO ’17, Berlin, Germany, 15–19 July 2017; pp. 641–648. [Google Scholar] [CrossRef]
- Chang, W.; Chyu, C. A Multi-Criteria Decision Making for the Unrelated Parallel Machines Scheduling Problem. J. Softw. Eng. Appl. 2009, 2, 323–329. [Google Scholar] [CrossRef]
- Shnits, B. Multi-Criteria Optimisation-Based Dynamic Scheduling for Controlling FMS. Int. J. Prod. Res. 2012, 50, 6111–6121. [Google Scholar] [CrossRef]
- Asanuma, K.; Nishi, T. Estimation of Weighting Factors for Multi-Objective Scheduling Problems using Input-Output Data. Trans. Inst. Syst. Control. Inf. Eng. 2022, 66, 1–9. [Google Scholar] [CrossRef]
- Hasani, K.; Kravchenko, S.A.; Werner, F. Simulated Annealing and Genetic Algorithms for the Two-Machine Scheduling Problem with a Single Server. Int. J. Prod. Res. 2014, 52, 3778–3792. [Google Scholar] [CrossRef]
- Lee, D.H.; Cao, Z.; Meng, Q. Scheduling of Two-Transtainer Systems for Loading Outbound Containers in Port Container Terminals with Simulated Annealing Algorithm. Int. J. Prod. Econ. 2007, 107, 115–124. [Google Scholar] [CrossRef]
- Li, X.; Gao, L. An Effective Hybrid Genetic Algorithm and Tabu Search for Flexible Job Shop Scheduling Problem. Int. J. Prod. Econ. 2016, 174, 93–110. [Google Scholar] [CrossRef]
- Togo, H.; Asanuma, K.; Nishi, T. Machine Learning and Inverse Optimization Approach for Model Identification of Scheduling Problems in Chemical Batch Plants. Comput.-Aided Chem. Eng. 2022, 49, 1711–1716. [Google Scholar] [CrossRef]
- Breiman, L. Random Forest. Kluwer Acad. Publ. 2001, 45, 5–32. [Google Scholar] [CrossRef]
- Chandrashekar, G.; Sahin, F. A survey on feature selection methods. Comput. Electr. Eng. 2014, 40, 16–28. [Google Scholar] [CrossRef]
- Zhang, H.; Ishikawa, M. A Solution to Inverse Optimization Problems by the Learning of Neural Networks. IEEJ Trans. Electr. Electron. Eng. 1997, 117-C, 985–991. [Google Scholar] [CrossRef]
- Tan, Y.; Delong, A.; Terekhov, D. Deep Inverse Optimization. In Proceedings of the International Conference on Integration of Constraint Programming, Artificial Intelligence and Operations Research, Thessaloniki, Greece, 4–7 June 2019; pp. 540–556. [Google Scholar] [CrossRef] [Green Version]
Weight of Job | Processing Time | Delivery Time | Setup Cost |
---|---|---|---|
1~5 | 50~500 | 50~200 | 10~1000 |
Processing Time | Delivery Time | Setup Cost |
---|---|---|
10~100 | 10~200 | 10~100 |
Features | 1. The values of the maximum completion time 2. The values of the sum of setup costs 3. The variance of completion time for each machine 4. Spearman’s rank correlation coefficients ) |
Only Objective Function Values | SFS | |
---|---|---|
MSE | 3.282 × 10−2 | 2.990 × 10−2 |
Features | 2. The values of the sum of setup costs 5. The values of the sum of delivery delay 7. The sum of completion time of each machine 6. The variance of delivery time setting for each machine 8. Spearman’s rank correlation coefficients ) |
Only Objective Function Values | SFS | |
---|---|---|
MSE | 2.875 × 10−3 | 2.871 × 10−3 |
Objective Function | I | II |
---|---|---|
Features | 2. The values of the sum of setup costs 3. The variance of completion time for each machine | 2. The values of the sum of setup costs 3. The variance of completion time for each machine 4. Spearman’s rank correlation coefficients ) 13. Spearman’s rank correlation coefficients ) 6. The variance of delivery time setting for each machine |
I (Only Objective Function Values) | I (SFS) | II (Only Objective Function Values) | II (SFS) | |
---|---|---|---|---|
MSE | 2.702 × 10−2 | 2.667 × 10−2 | 2.696 × 10−2 | 2.642 × 10−2 |
Small-Scale Problems (2 Objectives) | Large-Scale Problems (2 Objectives) | Small-Scale Problems (3 Objectives: I) | Small-Scale Problems (3 Objectives: II) | |
---|---|---|---|---|
MSE | 2.990 × 10−2 | 2.871 × 10−3 | 2.667 × 10−2 | 2.642 × 10−2 |
Small-Scale Problems (2 Objective) | Large-Scale Problems (2 Objective) | Small-Scale Problems (3 Objective: I) | Small-Scale Problems (3 Objective: II) | |
---|---|---|---|---|
MSE | 2.418 × 10−3 | 4.446 × 10−4 | 1.667 × 10−2 | 1.704 × 10−2 |
Machine Learning | Inverse Optimization | |
---|---|---|
Small-scale problems for 2 objectives | 2.990 × 10−2 | 2.418 × 10−3 |
Large-scale problems for 2 objectives | 2.871 × 10−3 | 4.446 × 10−4 |
Small-scale problems for 3 objectives (I) | 2.667 × 10−2 | 1.667 × 10−2 |
Small-scale problems for 3 objectives (II) | 2.642 × 10−2 | 1.704 × 10−2 |
Features | 2. The values of the sum of setup costs 5. The values of the sum of delivery delay 7. The sum of completion time of each machine |
Machine Learning | Inverse Optimization | ||
---|---|---|---|
Only Objective Function Values | SFS | ||
MSE | 5.020 × 10−2 | 4.796 × 10−2 | 2.345 × 10−2 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Togo, H.; Asanuma, K.; Nishi, T.; Liu, Z. Machine Learning and Inverse Optimization for Estimation of Weighting Factors in Multi-Objective Production Scheduling Problems. Appl. Sci. 2022, 12, 9472. https://doi.org/10.3390/app12199472
Togo H, Asanuma K, Nishi T, Liu Z. Machine Learning and Inverse Optimization for Estimation of Weighting Factors in Multi-Objective Production Scheduling Problems. Applied Sciences. 2022; 12(19):9472. https://doi.org/10.3390/app12199472
Chicago/Turabian StyleTogo, Hidetoshi, Kohei Asanuma, Tatsushi Nishi, and Ziang Liu. 2022. "Machine Learning and Inverse Optimization for Estimation of Weighting Factors in Multi-Objective Production Scheduling Problems" Applied Sciences 12, no. 19: 9472. https://doi.org/10.3390/app12199472
APA StyleTogo, H., Asanuma, K., Nishi, T., & Liu, Z. (2022). Machine Learning and Inverse Optimization for Estimation of Weighting Factors in Multi-Objective Production Scheduling Problems. Applied Sciences, 12(19), 9472. https://doi.org/10.3390/app12199472