Risk Assessment based on integrated fuzzy MEP methodology

Abstract

In the effort to overcome the limitations of some mathematical models to deal with real-world problems, evolutionary algorithms were developed as alternative optimization technique. Genetic Programming (GP) represents a branch of evolutionary programming, where encoding is performed by using an evolutionary algorithm and resulting in solutions that consist of computer programs. So far, genetic programming has been successfully implemented in various optimization, search and model approximation problems. As a subset of machine learning paradigm, GP, developed by Koza, uses genetic algorithms (GA) to automatically generate computer programs. Multi Expression Programming (MEP), a Genetic Programming variant, is used in developing models for characterization of system behavior by directly extracting knowledge from data. MEP is considered an efficient technique for solving complex problems, having a distinctive feature to store multiple solutions in a single chromosome. However, the decoding process remains at the same complexity level. MEP has a potentially wide range of applications. Real systems that exist in socio-economic environment are characterized by dynamic structure, reflected in nonlinearity, uncertainty and other inherent aspects. Thus, standard mathematical approach, relying on precise mathematical relations, has certain limitations in modeling complex systems. As an alternative, fuzzy mathematics is applied when modelling vague and complex relations and systems. The application of fuzzy systems theory is recommended in situations where data values and relations are uncertain and imprecise and their estimation relies on incomplete expert judgment. The principles of fuzzy mathematics have been extensively used in risk assessment. This paper proposes an integrated methodology for risk assessment that combines MEP and fuzzy mathematics.