Job Satisfaction Evaluation using Fuzzy Approach Khairul A. Rasmani and Nor Azni Shahari Faculty of Information Technology and Quantitative Sciences, Universiti Teknologi MARA, Malaysia {khairulanwar, norazni}@nsembilan.uitm.edu.my Abstract Likert-type scale that employs ordinal values to represent linguistics terms has been very popular in the studies on job satisfaction evaluation. In this work, it is argued that the use of ordinal values in Likert scale does not offer the best way in representing the linguistic terms. This paper proposes the use of fuzzy sets to represent linguistic terms in Likert-type scale and employs the technique using fuzzy conjoint method in job satisfaction evaluation. Experimental results show that the analysis using fuzzy conjoint method produced consistent result compared to the analysis using the percentage. However, the fuzzy membership values obtained from fuzzy conjoint method can be used to compare the decisions between criteria used to measure job satisfaction, and hence is very useful in providing additional information for decision-making. 1. Introduction Job satisfaction is an important aspect that determines the quality of academic staff in higher education. It illustrates enhancement of efficiency and performance of the institution. The outcome of a study on job satisfaction of academics staff also could also be very useful to provide information for the quality improvement process. In general, the instrument used for the evaluation of job satisfaction consists of several criteria that are very subjective. Typically, each criterion consists of several variables in which the respondents of the survey need to give their responses in the form of preferences or agreements such as ‘agree’ and ‘disagree’. Likert-type scale that employs ordinal values to represent linguistics terms has been very popular in the studies on job satisfaction. The discrete-ordinal values have been used for the analysis which includes the measurement of central tendency, correlation coefficient, regression, t-test, chi square test etc. For example in [4], five-point Likert response scale ranging from ‘1: very satisfied’, through ‘3:neutral’, to ‘5:very dissatisfied’ has been used in studies on job satisfaction among academic staff across eight nations including Australia, Germany, Hong Kong, Israel, Mexico, Sweden, UK and USA. In those studies, percentage and mean were used for the analysis. In [5], five-point Likert scale had been used in a study on the effect of change and transformation on academic staff and job satisfaction in South African University. The evaluation of job satisfaction was based on percentage. In [6], a ‘six-step’ Likert scale had been used in a study on job satisfaction of library staff at the University of North Carolina. The Likert scale consisted of ‘1:disagree very much’, ‘2:disagree moderately’, ‘3:disagree slightly’, ‘4:agree slightly’, ‘5:agree moderately’ and ‘6:agree very much’. In that study, the analysis was conducted using statistical mean and standard deviation. In [9], two types of Likert scale had been used, they were five- point scale ranging from ‘5:strongly agree’ to ‘1:strongly disagree’ and a seven-point Likert scale ranging from ‘0:never’ to ‘6:everyday’. The study which focused on job satisfaction and burnout among Greek early educators employed statistical mean, standard deviation and correlation for the analysis. It is observed that discrete-ordinal value is widely used to represent linguistic term in Likert-type scale. The use of percentage and statistical mean in particular are very popular methods for data analysis. For example in [4], statistical mean has been used to compare job satisfaction among academic staff in eight different countries. In the study, higher mean value indicates that the academic staffs are more satisfied to their jobs. However, the use of mean value obtained from Likert scale makes no sense because the ordinal value is just for coding. Furthermore, the interval between values is not interpretable in an ordinal measure. Thus, a mean value of ‘4.75’ from response to Likert scale ‘1:disagree’ to ‘5:agree’ does not mean that the outcome is ‘agree’. The use of discrete values in Likert scale is also inappropriate because the linguistic terms such as ‘agree’ and ‘disagree’ are measured as degrees of preferences or agreements which in nature are fuzzy terms. Thus, in this work, it is argued that the use of ordinal values in Likert scale does not offer the best way in representing the linguistic terms. As an alternative, fuzzy set offers a new way of representing the linguistic terms used in Likert scale. Moreover, the output from fuzzy inference in terms of membership value degree can be very useful to indicate the degree of preferences or agreements. The rest of this paper is organized as follows: Section 2 describes the background theory of fuzzy sets and fuzzy conjoint analysis method, Section 3 presents the setting for the experiment, the experimental results and discussion and finally, the conclusion and future work is outlined in Section 4.