Case-based Reasoning For Simulation modeling: Issues And Challenges Ming Zhou, PhD., Professor Center for Systems Modeling Simulation Indiana state University Terre Haute. IN 47809, USA Collaborating with Zhimin Chen, Ph. D, Professor School of Management, Shen Zhen University, PR China
Case-based Reasoning For Simulation Modeling: Issues And Challenges Ming Zhou, PhD., Professor Center for Systems Modeling & Simulation Indiana State University Terre Haute, IN 47809, USA Collaborating with Zhimin Chen, Ph.D., Professor School of Management, Shen Zhen University, PR. China
Reasons for simulation being underutilized Simulation modeling is a time-consuming and knowledge/information intense process Most models developed are customized? models that cannot be reused or easily adapted to other even Similar problems (Arons 1999, 2000; Zhou 2004) Conceptual modeling is a critical step that directly affects the quality and efficiency of Simulation project? but hardly supported with current technology, and still very difficult and ad-hoc process that depends on the skill and experience of individual modelers(Mclean, 2001; Robinson, 2004)
Reasons for simulation being underutilized • Simulation modeling is a time-consuming and knowledge/information intense process • Most models developed are customized “rigid” models that cannot be reused or easily adapted to other even similar problems (Arons 1999, 2000; Zhou 2004) • Conceptual modeling is a critical step that directly affects the quality and efficiency of simulation projects, but hardly supported with current technology, and still a very difficult and ad-hoc process that depends on the skill and experience of individual modelers (Mclean, 2001; Robinson, 2004)
Efforts made to address the difficulties Develop standard templates for specific classes of Simulation problems(pattern-based approach Develop modularized models or component-based modeling approach Develop standard interface that integrates simulation with other application systems Develop neutral data formats to facilitate model data transfer between different systems(to address interoperability)
Efforts made to address the difficulties • Develop standard templates for specific classes of simulation problems (pattern-based approach) • Develop modularized models or component-based modeling approach • Develop standard interface that integrates simulation with other application systems • Develop neutral data formats to facilitate model data transfer between different systems (to address interoperability)
Studies on knowledge-based simulation Develop extended programming languages,i.e genera al programming augmented with Simulation oriented language constructs Developing specialized simulation language based on a flow-chart type of logic Develop better interface to create a more interactive modeling environment (as opposed to batch? mode
Studies on knowledge-based simulation • Develop “extended programming languages”, i.e. general programming augmented with simulation oriented language constructs • Developing specialized simulation language based on a flow-chart type of logic • Develop better interface to create a more interactive modeling environment (as opposed to “batch” mode)
Continued development Object-oriented representation of simulation concepts has been emphasized since 1980s Case-based approach: adapt existing models for new applications(implemented models In software vendor industry, emphasis has been given to the development of high-level Simulators for special application systems
Continued development • Object-oriented representation of simulation concepts has been emphasized since 1980s • Case-based approach: adapt existing models for new applications (implemented models) • In software vendor industry, emphasis has been given to the development of high-level simulators for special application systems
Knowledge involved in simulation modeling Application Simulation Implementation Bounded by a context
Knowledge involved in simulation modeling Application Simulation Implementation Bounded by a context …
Two phases of modeling Involving different types of knowledge and different styles of reasoning Application Conceptualization Simulation Problem Problem Definition Definition Conceptual Implementation Implementation Simulation Simulation Model(CSM) Model (ISM)
Two phases of modeling • Involving different types of knowledge and different styles of reasoning… Application Problem Definition Simulation Problem Definition Conceptual Simulation Model (CSM) Implementation Simulation Model (ISM) Conceptualization Implementation
Deeper examination of SM process SM is an interactive decision-making process Problems in Sm are usually unstructured or semi-structured, i.e. the logic relations between decision factors is not well defined or clear SM is a knowledge/ information intense process Information and knowledge are used in a contextual manner, i.e. they are related to the unique structural and behavioral characteristics of a specific application. This context is very important in deriving solutions to similar problems but very difficult to store with conventional databases /information systems a great amount of knowledge/information is embedded among the solved simulation cases. A popular approach in practice is to develop a new simulation model by retrieving "old? models developed for similar past solved problems, and modifying the"old model to solve the new case
Deeper examination of SM process • SM is an interactive decision-making process • Problems in SM are usually unstructured or semi-structured, i.e. the logic relations between decision factors is not well defined or clear • SM is a knowledge/information intense process • Information and knowledge are used in a contextual manner, i.e. they are related to the unique structural and behavioral characteristics of a specific application. This context is very important in deriving solutions to similar problems but very difficult to store with conventional databases/information systems • A great amount of knowledge/information is embedded among the solved simulation cases. A popular approach in practice is to develop a new simulation model by retrieving “old” models developed for similar past solved problems, and modifying the “old” model to solve the new case
Problems with traditional KBs: rule-based expert systems(Watson, Leake, Bachant,.) Knowledge acquisition it is difficult to obtain generalized knowledge from SM processes due to the lack of basic understanding and unstructured nature problem domain. When problem domain is not well defined, the rules formulated are imperfect and produce unreliable solutions Knowledge elicitation it is difficult and laborious to extract empirical knowledge from human experts and formalize the knowledge into decision rules that can characterize the expert performance. However many rule-based systems assumed that expert knowledge is available and can be elicited and organized efficiently
Problems with traditional KBS: rule-based expert systems (Watson, Leake, Bachant, …) • Knowledge acquisition__ it is difficult to obtain generalized knowledge from SM processes due to the lack of basic understanding and unstructured nature of problem domain. When problem domain is not well defined, the rules formulated are imperfect and produce unreliable solutions • Knowledge elicitation__ it is difficult and laborious to extract empirical knowledge from human experts and formalize the knowledge into decision rules that can characterize the expert performance. However many rule-based systems assumed that expert knowledge is available and can be elicited and organized efficiently
Problems with traditional Kbs: rule based expert systems Knowledge maintenance in many applications rules are interrelated(e.g.chained with each other) and the number of rules required are unmanageably large Results interpretation in many domains, the inference process can become complex, and it is difficult for users to understand or verify the SOlutions suggested by rule-based reasoning
Problems with traditional KBS: rulebased expert systems • Knowledge maintenance__ in many applications, rules are interrelated (e.g. “chained” with each other) and the number of rules required are unmanageably large • Results interpretation__ in many domains, the inference process can become complex, and it is difficult for users to understand or verify the solutions suggested by rule-based reasoning