Keywords

OOBN; Pattern language; SDG; Bayesian networks;

Start Date

6-7-2022 12:00 PM

End Date

6-7-2022 12:20 PM

Abstract

Bayesian Networks (BN) are graph models that have been widely used in modeling applications where uncertainty treatment is a key consideration. When such applications strive to depict complex systems, an extension to the BN that utilizes the concepts of Object-Oriented programming (OOP) paradigm is adopted to reduce the model’s complexity and increase its understandability and modularity, namely Object-Oriented Bayesian Networks (OOBN). Although OOBN serve as a versatile framework that is apt for addressing a wide range of the challenges pertinent to developing complex interdisciplinary models, the development of OOBN can be a challenging task that comprises multiple critical design choices affecting the output model, e.g. determining OOBN classes and structure. This paper proposes a "Pattern language (PL)" for building OOBN models. The use of patterns as reusable transferable containers of knowledge has been promoted by many researchers in different fields, where a pattern, in this context, describes the core of a proven solution to a recurring design problem. While applying patterns may support modelers by providing them with solutions to the problems they will likely face during model building, patterns fall short in describing how a whole system can be designed or how cross-domain knowledge can be integrated to solve complex problems. A pattern language, however, coherently organizes patterns describing the relationships and dependencies between them so they can be used to develop complex designs. The OOBNPL presented herein explains how an assembly of patterns compiled from BN, OOP, and modeling knowledge domains work together to enable modelers to produce effective probabilistic component-based models. In addition to increasing the efficiency of the modeling process, applying the proposed OOBN-PL will help increase the OOBN models modularity, composability, and reusability. The use of the proposed OOBN-PL is demonstrated through a use case for building of a model that attempt to capture the impacts of SDG interconnections on the long-term sustainability and resilience of nations.

Stream and Session

false

Share

COinS
 
Jul 6th, 12:00 PM Jul 6th, 12:20 PM

The development of OOBN Pattern Language: Design approach and case study in SDG modelling

Bayesian Networks (BN) are graph models that have been widely used in modeling applications where uncertainty treatment is a key consideration. When such applications strive to depict complex systems, an extension to the BN that utilizes the concepts of Object-Oriented programming (OOP) paradigm is adopted to reduce the model’s complexity and increase its understandability and modularity, namely Object-Oriented Bayesian Networks (OOBN). Although OOBN serve as a versatile framework that is apt for addressing a wide range of the challenges pertinent to developing complex interdisciplinary models, the development of OOBN can be a challenging task that comprises multiple critical design choices affecting the output model, e.g. determining OOBN classes and structure. This paper proposes a "Pattern language (PL)" for building OOBN models. The use of patterns as reusable transferable containers of knowledge has been promoted by many researchers in different fields, where a pattern, in this context, describes the core of a proven solution to a recurring design problem. While applying patterns may support modelers by providing them with solutions to the problems they will likely face during model building, patterns fall short in describing how a whole system can be designed or how cross-domain knowledge can be integrated to solve complex problems. A pattern language, however, coherently organizes patterns describing the relationships and dependencies between them so they can be used to develop complex designs. The OOBNPL presented herein explains how an assembly of patterns compiled from BN, OOP, and modeling knowledge domains work together to enable modelers to produce effective probabilistic component-based models. In addition to increasing the efficiency of the modeling process, applying the proposed OOBN-PL will help increase the OOBN models modularity, composability, and reusability. The use of the proposed OOBN-PL is demonstrated through a use case for building of a model that attempt to capture the impacts of SDG interconnections on the long-term sustainability and resilience of nations.