Development of a tool for knowledge-based linear programming formulation for production management domain

Abstract

Linear Programming (LP) is a very useful mathematical programming tool for solving various optimization problems in many application areas. However in the life cycle of LP model development, there has been little support in the modeling stage. The necessary aids in modeling include not only the syntactic knowledge of LP structure, but also semantic knowledge of application domain. In this thesis, we propose a method that the 1P modeling process can be supported by a knowledge base. Modeling knowledge is thus represented in frames with the additional features like object concept. Frame representation can describe the model in multiple point of views, and can adopt to the dynamic change of the modeling situations. The dialog and explanation can also be generated by the modeling knowledge. The required data can be retrieved by tight coupling with existing DBMS using the syntaxdriven interfaces. The propose knowledge-based modeling method can be used as a stand alone Decision Support System (DSS) and as a modeling part of the Post-Model Analysis (PMA) environment. We have developed a prototype, KLIPF (Knowledge-based LInear Programming Formulation), to implement our proposed method for the domain of production management. KLIPF is implemented using LISP on IBM-PC/AT.