New research aids decision-makers in the analysis of optimization results by giving insights into the otherwise "black-box" optimization process. An important outcome of this research is an openly available interactive knowledge discovery and decision support tool called Mimer.
Making a decision based on data can be challenging in many real-world situations. It may be because there are several variables to be analyzed, or because there are multiple criteria to be considered, each of which points to a different solution.
This becomes even more challenging when more than one person has a say in the final decision, and they all have their preferences. How then, can a business or an organization make complex decisions that are grounded in data? The key is to transform data into information, and that information into knowledge.
“Our research uses data mining and visual analytics to facilitate decision-makers in discovering, visualizing, interacting with and sharing knowledge. The focus is to make this transformation from data to knowledge easy and efficient so that effective data-driven decisions can be made” says Sunith Bandaru, associate professor of production engineering.
Learning from the optimization process
The research specifically considers data generated by optimization algorithms when solving real-world problems involving multiple criteria. These algorithms use principles of natural evolution proposed by Charles Darwin to “evolve” a population of “individuals” over several “generations”.
Each individual is a possible solution to the problem. By starting with a population of random solutions, these algorithms are capable of generating several optimal solutions. Thus, decision-makers have many alternatives to choose from depending on how much importance they place on each criterion.
The current research shows how important knowledge about the problem can be gained from optimization data, and how this knowledge can be visualized. For example, the decision-makers easily learn what is the difference between good solutions and bad solutions, or how the preferences of one decision-maker differ from those of others.
Speeding up the optimization process
“This research offers a way for decision makers to gain knowledge about the structure of multi-objective optimization problems that can be used both to update existing or design future problem formulation or to help in the optimization process itself with knowledge-driven optimization, by incorporating the knowledge into the optimization algorithm” says Henrik Smedberg, PhD student in Informatics.
A snapshot of Mimer - an openly available interactive knowledge discovery and decision support tool. Try it now at https://assar.his.se/mimer/html/.
The research can be used by anyone performing multi-objective optimization and has the potential of both speeding up the optimization process and also bringing insight to decision-makers that can benefit the design of future optimization cases.
Try Mimer – A decision support tool
Henrik Smedberg and Sunith Bandaru say that one important outcome of this research is an openly available interactive knowledge discovery and decision support tool called Mimer. Try it now at https://assar.his.se/mimer/html/
The tool is very easy to use and on the tutorials page, you will find several examples showcasing its various features.
More about this research
The research presented in the paper was conducted with optimization data, and researchers within VF-KDO have published other papers where Mimer has been used to obtain knowledge about multi-criteria optimization problems from the manufacturing industry, see the list below.
However, note that Mimer as a knowledge discovery tool is not limited to optimization problems. Since it implements data mining methods in the backend, it can be used with any numerical data where there are some inputs and outputs, and the user is interested in finding hidden patterns of inputs that lead to certain values of output. Such knowledge often helps explain interesting phenomena in complex real-world systems and thus can give deep insights.
Industry cases where Mimer has been used
A. Iriondo Pascual, H. Smedberg, D. Högberg, A. Syberfeldt, and D. Lämkull, “Enabling knowledge discovery in multi-objective optimizations of worker well-being and productivity,” Sustainability, vol. 14, no. 9, p. 4894, 2022.
C. A. B. Diaz, H. Smedberg, S. Bandaru, and A. H. Ng, “Enabling knowledge discovery from simulation-based multi-objective optimization in reconfigurable manufacturing systems,” in 2022 Winter Simulation Conference (WSC), 2022, pp. 1794–1805.
C. A. Barrera-Diaz, A. Nourmohammdi, H. Smedberg, T. Aslam, and A. H. Ng, “An enhanced simulation-based multi-objective optimization approach with knowledge discovery for reconfigurable manufacturing systems,” arXiv preprint arXiv:2212.00581, 2022.
H. Smedberg, C. A. Barrera-Diaz, A. Nourmohammadi, S. Bandaru, and A. H. Ng, “Knowledge-driven multi-objective optimization for reconfigurable manufacturing systems,” Mathematical and Computational Applications, vol. 27, no. 6, p. 106, 2022.