FIE: a generic fuzzy decision making tool with an example of CRM analysis.

• Author: Kouatli, Issam

1. INTRODUCTION

   Management of information systems today are much more complex due to the variety of socio-technical elements introduced as the system evolve. Fuzzy logic (Zadeh 1965 & 1973) [1,2] is one type of theory that can be used to help achieving a decision given uncertain (fuzzy) inputs and as such it can provide a translation of the qualitative abilities of the human brain into quantitative functions. Uncertainty in the decision making is becoming more and more needed as most decisions are becoming too fluid and unstructured. A tool that can help managers to take a decision under uncertainty is becoming highly demanded. Many systems uses fuzzy inference engine as an embedded features of their perspective systems, for example, Cheng et al.[3] proposed an e-marketplace negotiation system that is based upon fuzzy inference system. Dweiri et.al.[4] proposed to use fuzzy decision making systems to measure the efficiency based on three criteria, project cost, project time and project quality. Most mathematical modeling of decision support system may lack the existence of accuracy element due to the fact that either because the variables are actually "fuzzy" (not accurate) or due to the fact that the human individual's interpretation of the inputs may vary. This fact has led a number of researchers in the field to adopt the mixture of accurate mathematical formulae with some elements of fuzziness. Jimenez et.al.[5] for example has proposed a method for solving linear programming problems using fuzzy parameters and Wang et al. [6] proposed a fuzzy modification to AHP (Analytical Hierarchy processing) to aid decision maker in optimization of maintenance strategies.

   This paper introduces a new user-friendly fuzzy inference tool that can be utilized by managers or by researchers to aid them in decision making without the need to understand the mechanism of fuzzy logic/system. It is basically an inference engine (shell) using fuzzy logic and hence termed as "FIE" (Fuzzy Inference Engine). FIE was developed on the basis of the concept of Fuzzimetric Arcs for fuzzy set choice and selection introduced by Kouatli et al.[7] as well as then concept of multivariable system introduced by Kouatli [8]. Methodology of definition and selection of fuzzy sets (Fuzzy variables) using the concept of Fuzzimetric Arcs in conjunction with Genetic algorithm was also proposed by Kouatli [9]. The structure of Fuzzy Inference Engine (FIE) based on the above work [7,8,9] which will be reviewed and the mechanism/components of FIE will be illustrated. JAVA language was used to implement the concept of FIE. An example of using FIE to CRM performance analysis was used as a vehicle to demonstrate FIE Operation.

2. STRUCTURE OF FIE

   FIE structure was based on two concepts, the fuzzimetric arcs and Fuzzy Multivartiable structure. The fuzzimetric arcs is a fuzzification mechanism to define and select the most appropriate fuzzy sets/variables. The second concept related to a simplified approach of dealing with multivariable rule-set that define the operation of decision making system.

   2.1 CONCEPT OF FUZZIMETRIC ARCS

   Fuzzimetric arcs concept is a methodology for a simpler universe partitioning technique as well as a methodology for the selection of the optimum shape of the fuzzy sets for the fuzzy variables to be developed. Each arc is mainly composed of three quarters of a trigonometric circle and which has a radius of absolute value of unity. Then the main fuzzy variables zero, Small, Medium and Big of any system may be represented on this arc, which may encompass positive and/or negative values forming the process of partitioning. Hence, for positive values, the members of the universe that belong to the fuzzy variable Positive Zero (PO) for example, would carry a membership value of unity (cos 0 = sin ([pi]/2 - 0) = 1) for the member zero which decreases gradually for the rest of the members until the start of the second quarter of the arc is reached (Fig. 1).

   Similarly, the Positive Small (PS) fuzzy variable should carry a non-zero membership value for any member that lies in the range of 0[degrees] to 180[degrees] on the arc where a membership value of unity (sin [pi]/2) should be at the start of the second quarter of the arc. The members that belong to the fuzzy variable Positive Medium (PM) have a non-zero membership value in the range of 90[degrees] to 270[degrees] on the arc. The final fuzzy variable Positive Big (PB)...