.png){kind=icon}
.png){kind=icon}
Return
Data-driven optimization model customization
.png){kind=icon}
PRE
AI
DOI：10.1016/j.ejor.2020.05.010.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
When embedded in software-based decision support systems, optimization models can greatly improve organizational planning. In many industries, there are classical models that capture the fundamentals of general planning decisions (e.g., designing a delivery route). However, these models are generic and often require customization to truly reflect the realities of specific operational settings. Yet, such customization can be an expensive and time-consuming process. At the same time, popular cloud computing software platforms such as Software as a Service (SaaS) are not amenable to customized software applications. We present a framework that has the potential to autonomously customize optimization models by learning mathematical representations of customer-specific business rules from historical data derived from model solutions and implemented plans. Because of the wide-spread use in practice of mixed integer linear programs (MILP) and the power of MILP solvers, the framework is designed for MILP models. It uses a common mathematical representation for different optimization models and business rules, which it encodes in a standard data structure. As a result, a software provider employing this framework can develop and maintain a single code-base while meeting the needs of different customers. We assess the effectiveness of this framework on multiple classical MILPs used in the planning of logistics and supply chain operations and with different business rules that must be observed by implementable plans. Computational experiments based on synthetic data indicate that solutions to the customized optimization models produced by the framework are regularly of high-quality. (C) 2020 Elsevier B.V. All rights reserved.
Keywords:
Decision support systems
Statistical learning
Mixed integer linear programming
Optimization modeling
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
6
Papers:
2.2W
Citations:
6.4W