Scholarly article on topic 'Analysis of the Job Satisfaction Index Problem by Using Fuzzy Inference'

Analysis of the Job Satisfaction Index Problem by Using Fuzzy Inference Academic research paper on "Economics and business"

CC BY-NC-ND
0
0
Share paper
Academic journal
Procedia Computer Science
OECD Field of science
Keywords
{"job satisfaction" / "fuzzy logic" / "possibility measure" / "expert system" / "fuzzy inference" / "Minnesota Satisfaction Questionnaire" / "quality measure- degree of truth"}

Abstract of research paper on Economics and business, author of scientific article — T. Saner, Latafat A. Gardashova, Rashad A. Allahverdiyev, Serife Z. Eyupoglu

Abstract In this study we propose a fuzzy rule-based algorithm to evaluate job-satisfaction in an organization. We collected the effective factors/job facets of job satisfaction through interviews. Through analyzing the interview results we compose fuzzy rules. By using the obtained rules, the value of job satisfaction is computed using the expert system shell ESPLAN. The basic advantage of the used approach is being able to operate with imperfect information for the evaluation of job satisfaction by using fuzzy logic.

Academic research paper on topic "Analysis of the Job Satisfaction Index Problem by Using Fuzzy Inference"

CrossMark

Available online at www.sciencedirect.com

ScienceDirect

Procedía Computer Science 102 (2016) 45 - 50

12th International Conference on Application of Fuzzy Systems and Soft Computing, ICAFS 2016, 29-30

August 2016, Vienna, Austria

Analysis of the job satisfaction index problem by using fuzzy inference

T.Saner a, Latafat A. Gardashova b*, Rashad A.Allahverdiyev c , Serife Z. Eyupoglu d

a School of Tourism and Hotel Management, Near East University, P.O.Box:99I38, Nicosia, North Cyprus, Mersin 10 Turkey b'*Azerbaijan State Oil and Industry University, Department of Computer-Aided Control Systems, Baku, Azerbaijan c Azerbaijan State Oil and Industry University, BA Programs of ASOIU, Baku, Azerbaijan, _Department of Business Administration, Near East University, P.O.Box:99I38, Nicosia, North Cyprus, Mersin 10 Turkey_

Abstract

In this study we propose a fuzzy rule-based algorithm to evaluate job-satisfaction in an organization. We collected the effective factors/job facets of job satisfaction through interviews. Through analyzing the interview results we compose fuzzy rules. By using the obtained rules, the value of job satisfaction is computed using the expert system shell ESPLAN. The basic advantage of the used approach is being able to operate with imperfect information for the evaluation of job satisfaction by using fuzzy logic.

© 2016 The Authors. PublishedbyElsevierB.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of the Organizing Committee of ICAFS 2016

Keywordsjob satisfaction; fuzzy logic; possibility measure; expert system; fuzzy inference; Minnesota Satisfaction Questionnaire; quality measure- degree of truth

1. Introduction

There is a vast range of studies devoted to job satisfaction in the existing literature. Job satisfaction theories have a strong overlap with theories explaining human motivation. The most common and prominent theories in this area include Maslow's Needs Hierarchy Theory1, Herzberg's2 Motivator-hygiene Theory, the Job Characteristics Model3, and the Dispositional Approach4. These theories are popular in the literature related to human motivation5,8. Some determinants of job satisfaction are analyzed in9,12. In13aneffective approach to job satisfaction is described. Job satisfaction indicators and their featuresare described in14.

Job satisfaction is not only about how much an employee enjoys work. Taber and Alliger15 analyzed other types of measures such as level of concentration required for the job, level of supervision, and task importance. This study demonstrates that the accumulating enjoyment of work tasks add up to an overall job satisfaction.

Some factors of job satisfaction may be ranked as more important than others, depending on each worker's needs and personal and professional goals. To create a benchmark for measuring and ultimately creating job satisfaction, managers in an organization can employ proven test methods such as the Job Descriptive Index (JDI) or the Minnesota Satisfaction Questionnaire (MSQ)16. These assessments help management define job satisfaction adequately.

Five important factors/job facets can be used to measure and influence job satisfaction in the test methods17: 1. Pay or total compensation

2. The work itself (i.e., job specifics such as projects, responsibilities)

3. Promotion opportunities (i.e., expanded responsibilities, more prestigious title)

4. Relationship with supervisor

5. Interaction and work relationship with coworkers.

'Corresponding author. Tel.: +994505840901 E-mail address: latsham@yandex.ru

1877-0509 © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of the Organizing Committee of ICAFS 2016 doi:10.1016/j.procs.2016.09.368

In18, the authors propose a fuzzy rule-based algorithm to evaluate the job satisfaction in an organization. First, they collect the effective factors/job facets of job satisfaction through interviews. After analyzing the interview results, they propose questionnaires with respect to categories obtained from interviews. Due to qualitative aspect of satisfaction, they use linguistic choices in the questionnaires. While it is hard to disseminate questionnaires to all employees, sampling is performed based on STRATA technique. The results are used to compose fuzzy rules. After defuzzification of the rules and computing the distance from ideal status, the gaps are determined. The gaps are fulfilled using improvement strategies. Next, they give a brief description of STRATA sampling technique and fuzzy logic. Fuzzy logic is capable of treating this dynamic performance criterion in the uncertain and qualitative environment.

Authors19 examine how individuals "determine" their job satisfaction based on changes in situational factors. A simulation model, using fuzzy set theory and system dynamics, is used.As Piegat20 stated "information obtained from people is usually of less precision (large granularity), while information delivered by measuring devices is of higher precision (small granularity)". For the model, the requiredinformation is obtained from people. It measures subjective features of work, consequently making fuzzy set theory a highly applicable technique to evaluate the features. The estimation of an individual's input-output ratio and the impacts of the input-output ratio on changes to the individual's satisfaction level are evaluated using fuzzy set theory. Fuzzy logic is used to get an approximate answer when no exact answer is possible.

The purpose of this study is to determine the level of employee job satisfaction throughthe use of the Minnesota Satisfaction Questionnaire (MSQ) based production rules and the fuzzy expert system shell ESPLAN21.

The paper is organized as follows. Section 2 discusses description of fuzzy if-then rules and fuzzy inference algorithm. The statement of the problem is described in section 3. Theresults of computer simulation are described in Section 4 and Section 5 concludes the study.

2. Description of fuzzy if-then rules and fuzzy inference algorithm

Knowledge in if-then rules based systems can be described in different ways. Some of the post-modern techniques for representation of knowledge include logical calculus and a structured model. This work is devoted to the rule-based system oriented approaches of knowledge representation. A rule-based system consists of three main parts:1) a set of if-then rules, 2) dynamic database, called the working-storage, 3) control interpreter, which interprets the database using the set of rules. The rule-based system has a wide class of applications in decision making problems, planning problems, business problems, technical problems, and in social sciencessuch as psychology and medicine21,24.

The shell of ESPLAN provides the following basic abilities:development of expert systems for various applications; building module-oriented structures and knowledge bases segmentation; representation of fuzzy values; compositional inference with possibility measures; arithmetic operations with fuzzy numbers; realization of simple user-machine dialogue (execution of queries) by using special functions; the use of a confidence degree for any rule (in percent); call of external programs; data interchange using file system..

The mathematical description of knowledge in the knowledge base is based on fuzzy interpretation of antecedents and consequents in if-then rules.

For the knowledge representation the antecedent of each rule contains a conjunction of logical connectives like <linguistic

value> named elementary antecedent.

The consequent of the rule is a list of imperatives, among which may be some operator-functions (i.e. input and output of objects' values, operations with segments of a knowledge base, etc). Each rule may be characterized with a confidence degree

Each linguistic value has a corresponding membership function. The subsystem of fuzzy arithmetic and linguistic values processing provides automatic interpretation of linguistic values like "approximately A", "less than A", "more than A", middle", "much", "high", "low", "near...", "from ... to..." and so on; i.e. for each linguistic value this subsystem automatically computes parameters of membership functions using universes of corresponding variable. The value of linguistic variable are trapezoidal fuzzy numbers, which is described as21:

less than A : (0,I, A - Z, Z) approximately A: (Z, A, A, Z) more thanA: (Z, A + Z, S,0) neutral: (Z,I, +2*Z,I + 3*Z,Z)much: (Z,S - Z,S,0) etc,

where I and S are respectively minimum and maximum values of universe, Z=(S-I)/5.

The user of the system may define new linguistic values, modify built-in ones and explicitly prescribe a membership function in any place where linguistic values are useful21. The fuzzy if-then rules have the following form: R k: IF X is Ak1 and x2 is At 2 and... and xm is Am THEN ut1 is Bk1 and ut 2 is Bt 2 and... and uk1 is Bk1 , k = 1, K where xt, i = 1, m and Uj, j = 1, l are total input and local output variables , Au,BkJ are fuzzy sets, and k is the number of rules. Note, that inputs x1, x2,..., xm may be crisp or fuzzy variables.

Efficiency of the inference engine considerably depends on the knowledge base internal structure. Theinference mechanism acts as follows. First, current values of objects are given (initial data). Then all rules antecedents of whichoverlapswith these current values are chosen from the knowledge base. For these rules, the truth degrees of the rules are computed (in other words,

Cf e[0,100].

the system estimates the truth degree of the fact that current values of objects correspond to values fixed in antecedents). If the truth degree exceeds some threshold then imperatives from consequents of a rule are executed. The assigned value of the object is also complemented by a numeric confidence degree.

A truth degree of a rule's antecedent is calculated according to the following algorithm21. 1.First the objects are evaluated, i.e. every w, object has appropriate value defined as (vt, cfk) ,where vk is linguistic value,

cf e [0,1000] is confidence degree of the value vk. Then it is needed to compute:

r k = Poss(vk / aJk j ■ cfk,, if the sign is "="or rt = -Poss{vk | jjcfk, , if the sign is 'V.

Poss is defined as2Poss(vl a) = m„ax min(<"i (u), ^ (u)) e [0,1] TJ = min (rk)

aJt - current linguistic value (j is index of the rule, k is index of fuzzy relation)

2.For each rule, calculate RJ = |min rJk J * CFJ / 100 ,

where CF is the confidence degree of the rule .

The user or the creator of the rule defines the firing level ( n ) and Rj > n is checked. If the condition holds true, then the consequent part of rule is calculated.

3.The evaluated wt objects have St values21: wt, (v1, cf}),...,..., jvf', cfS'^ The consequents of rules areaggregated into the average21:

tv," ■ cf,"

,, _ n=1_

Hcf" "=1

IF aj AND x2 = aJ2 AND .„ THEN y1 = bj AND y2 = b^ AND... IF.„ THEN Y = AVRG(yl) AND Y2 = AVRG(y2) AND...

This model has a built-in function AVRG which calculates the average value. This function simplifies the implementation of compositional inference with possibility measures. As a possibility measure, here a confidence degree is used. So, the compositional relation is given as a set of rules like

IF x = AJ AND x2 = Aj ... THEN y1 = Bj AND y2 = BJ AND,

where J is a number of a rule. After all these rules have been executed (with different truth degrees) the next rule (rules) ought to be executed:

IF THEN Y = AVRG(yl) AND Y2 = AVRG(y2) AND... Using this model one may construct hypotheses. Such system contains the rules: IF < condition j > THEN X = A, CONFIDENCE cf,

Here "X = A," is a hypothesis that the object X takes the value A, . Using some preliminary information, this system generates elements X ,Rj) ,where Rj is a truth degree of J-th rule. In order to account the hypothesis (i.e. to estimate the truth degree that X takes the value A, the recurrent Bayes-Shortliffe formulaf generalized for the case of fuzzy hypotheses, is

used21:

P0 = 0

pj = PJ + cfjPoss {A>/ A j^1 - J

This formula is realized as a built-in function BS :IF END THEN P = BS (x, A ) .

3. Statement of the problem

Defining overall job satisfaction is a very important problem. The basic problem is to evaluate overall satisfaction of respondents by using job facets.For determining overallsatisfaction from evaluation of job facets,we use fuzzy rules.The overall job satisfaction denoted y is a compound index built from twenty components each of which is assessed by an expert judgement. The twenty components are: x,-Activity, x2 -Independence, x3-Variety x4-Social status, x5-Supervision-human relations, x6 -Supervision-technical, x7 - Moral values, x-Security, x9 -Social service, xi0- Authority, xn -Ability, x12-Company policies and practices, x13 -Compensation , x14 -Advancement, x15 - Responsibility, x16 -Creativity , x17 -Working conditions, x18 -Co-workers, x19 -Recognition, x20 -Achievement.

Using the above mentioned parameters, the overall job satisfaction model can be expressed as:

IF x = " very .satisfied " AND x2 = " very .satisfied " AND x3 = " quite satisfied " AND x4 = " less satisfied " AND x5 = " quite satisfied " AND x6 = " quite satisfied " AND x7 = " very satisfied " AND x8 = " satisfied " AND x9 = " very satisfied " AND x10 = "quite satisfied" AND x11 = "very satisfied" AND x12 ="satisfied" AND x13 = "very satisfied" AND x14 = "very satisfied" AND x15 ="very satisfied" AND x16 ="very satisfied" AND x17 = "very satisfied" AND x18 ="quite satisfied" AND x19 = "satisfied" AND x20 = "very satisfied" THEN y="satisfied";

IF x ="very satisfied" AND x2 ="satisfied" AND x3 = "very satisfied" AND x4 = "very satisfied" AND x5 = "very satisfied" AND x6 ="satisfied" AND x7 ="quite satisfied" AND x = "quite satisfied" AND x9 = "very satisfied" AND x10 ="satisfied" AND x11 = "very satisfied" AND x12 = "quite satisfied" AND x13 ="satisfied" AND x4 = "very satisfied" AND x15 = "very satisfied" AND x16 = "very satisfied" AND x17 = "satisfied" AND x18 ="quite satisfied" AND x19 = "very satisfied" AND x20 = "very satisfied" THEN y="satisfied"; ...... (2)

IF x = "quite satisfied" AND x2 = "quite satisfied" AND x3 = "quite satisfied" AND x4 = "quite satisfied" AND x5 = "satisfied" AND x6 ="satisfied" AND x7 = "quite satisfied" AND x8 = "quite satisfied" AND x9 = "quite satisfied" AND x10 = "quite satisfied" AND x11 ="quite satisfied" AND x12 ="less satisfied" AND x13 ="less satisfied" AND x14 ="quite satisfied" AND x15 ="satisfied" AND x16 = "quite satisfied" AND x17 = "less satisfied" AND x8 ="satisfied" AND x19 = "quite satisfied" AND x20 = "quite satisfied" THEN y="quite satisfied".

These rules have been extracted from experts' knowledgebased on interviews conducted by us.The trapezoidal fuzzy numbers describing the used linguistic terms are given below:

unsatisfied or less than A: ( 0, I, A - Z, Z) ;less satisfied or approximately A: ( Z, A , A , Z );

very satisfied or more than a = ( Z , A + Z , S , 0 ) ;quite satisfied or neutral = ( Z , I , + 2 * Z , I + 3* Z , Z ) ;

satisfied or much = ( Z , S - Z , S , 0 )

where I and S- respectively minimum and maximum value of universe, Z=(S-I)/5. Graphical representation ofthetrapezoidal fuzzy numbers is given in Fig. 1

2 2.5 3 3.5

Fig. 1 Linguistic terms for "job satisfaction"

In the expert system shell ESPLAN other linguistic terms can be used such asafew, average, more than A, less than A, approximately A, from A to B, strict more than A, strict Less than A and etc. For every linguistic value,ESPLAN

automatically calculates fuzzy number by using the universe .For instance, object= "activity", I=minimum=1, S=maximum=5, linguistic term="quite satisfied": quite satisfied or neutral: (Z,I + 2*Z,I + 3*Z,Z)

Our aim is to define the level of overall job satisfaction level using twenty job facets represented by fuzzy linguistic terms. 4. Computer simulation

The above mentioned model is implemented by using the fuzzy expert system ESPLAN and different tests are performed.Different current information in tests is used.

Test 1: IF x is quite satisfied AND x2 is quite satisfied AND Xjis less satisfied AND x4 is less satisfied AND x5is unsatisfied AND x6 is unsatisfied AND x7 satisfied AND x8 is quite satisfied AND x9 is less satisfied AND x10 is quite satisfied AND x11 is less satisfied AND x12 is quite satisfied AND x13 is quite satisfied AND x14 is quite satisfied AND x5 is unsatisfied AND x16 -is less satisfied AND x17 is quite satisfied AND x18 is very satisfied AND x19 is very satisfied AND x20 is satisfied THEN overall job satisfaction=?

Test 2:IF x, is satisfied AND x2 is quite satisfied AND xjis satisfied AND x4 is less satisfied AND x5is quite satisfied AND x6 is very satisfied AND x7 is satisfied AND x is quite satisfied AND x9 is quite satisfied AND x10 is satisfied AND x11 is quite satisfied AND x12 is satisfied AND x13 is quite satisfied AND x14 is quite satisfied AND x15 is quite satisfied AND x16-is satisfied AND x17 is satisfied AND x18 is quite satisfied AND x19 is quite satisfied AND x20 is quite satisfied THEN overall job satisfaction=?

Test 3:IF x is satisfied AND x2 is very satisfied AND xjis quite satisfied ANDx4 is very satisfied AND x5is satisfied AND x6 is satisfied AND x7 is quite satisfied ANDx8 is satisfied AND x9 is satisfied AND x10 is very satisfied AND Xi is satisfied AND x12 is quite satisfied AND x13 is satisfied AND x14 is very satisfied AND x15 is satisfied AND x16 is very satisfied AND x17 is satisfied AND x8 is satisfied AND x9 is very satisfied AND x20 is satisfied THEN overall job satisfaction=? FOR TEST1. ANSWER:

EXPERT system shell ESPLAN's decision is "Overall job satisfaction is LESS SATISFIED"

FOR TEST2.

ANSWER:

EXPERT system shell ESPLAN's decisionis "Overall job satisfaction is QUITE SATISFIED"

FOR TEST3.

ANSWER:

EXPERT system shell ESPLAN decision is "Overall job satisfaction is SATISFIED"

Fragment of computer simulation is given below.

Activity=Quite satisfied Independence=Quite satisfied Variety=Less satisfied Social status=Less satisfied Supervision-human relations=U"satisfied Supervision-technical=U"satisfied Moral values=Satisfied Security= Quite satisfied Social service=Less satisfied Authority=Quite satisfied Ability=Less satisfied

Company policies and practices= Quite satisfied Compensation= Quite satisfied Advancement= Quite satisfied Responsibility=U"satisfied Creativity=Less satisfied Working conditions=Quite satisfied Co-workers=Very satisfied Recognition=Very satisfied Achievement.=Satisfied

OVERALL JOB SATISFACTION is LESS THAN SATISFIED 5. Conclusion

In this paper for the evaluation of an overall job satisfaction index a fuzzy rule-base method is used. By using the Minnesota Satisfaction Questionnaire, values of basic determinants of respondents were determined. The fuzzy rules extracted by using interviewswere performedin the expert system shell ESPLAN and different tests were performed. The obtained results of job satisfaction evaluation on the bases of real data show validity and efficiency of the suggested approach.

References

1. Maslow A. Motivation and Personality. New York, Harper, 1995.

2. Herzberg F. Work and the Nature of Man. Cleveland, World Publishing Company, 1966.

3. Hackman JR, Oldham GR. Development of the Job Diagnostic Survey. .Journal of Applied Psychology, 60, 1975, pp. 161.

4. Judge TA, Lanen RJ. Dispositional affect and job satisfaction: A review and theoretical extension.Organizational Behavior and Human Decision Processes, 8,

6, 2001, pp. 67.

5. Furnham A, Eracleous A. Chamorro-Premuzic T. Personality, motivation and job satisfaction: Hertzberg meets the Big Five.Journal ofManagerial

Psychology, 24, 8, 2009, pp. 765.

6. Mitchell OS, Levine PB, Pozzebon S. Aging, job satisfaction, and job performance (CAHRS Working Paper #90-02). Ithaca, New York: Cornell University,

School of Industrial and Labor Relations, Center for Advanced Human Resource Studies. Retrieved 3 March 2013

7. Nagy MS.Using a single-item approach to measure facet job satisfaction. Journal of Occupational and Organisational Psychology, 75, 1, 2002, pp. 77.

8.Astrauskaite M, Vaitkevicius R, Perminas A. Job Satisfaction Survey: A confirmatory factor analysis based on secondary school teachers' sample. International Journal of Business and Management, 6, 5, 2011, pp. 41

9. Ewen RB.Some determinants of job satisfaction: A study of the generalisability of Herzberg's theory. Journal of Applied Psychology, 48, 1964, pp. 161.

10. Manisera M., Dusseldorp E, van der Kooij AJ. Component structure ofjob satisfaction based on Herzberg's theory, Rapporti di Ricerca del Dipartimento di

Metodi Quantitativi, Brescia University, Working Paper 253, 2005.

11. Behson SJ. 'Using relative weights to reanalyse 'settled' areas of organisational behaviour research: The job characteristics model and organisational justice',

International Journal of Management and Information Systems, 14, 5, 2010, pp. 43.

12. Fried Y, Ferris GR. 'The validity of the Job Characteristics Model: A review and meta-analysis', Personnel Psychology, 40, 1987, pp. 287.

13. Gerhart B. The (affective) dispositional approach to job satisfaction: Sorting out the policy implications. Journal of Organisational Behaviour, 26, 2005, pp.

14. Stanley E. Seashore, Thomas D. Taber. Job Satisfaction Indicators and Their Correlates. American Behavioral scientist, Vol. 18 No.3,January/February,

1975. Sage Publications, Inc. pp. 333-368

15. Taber TD, Alliger GM. A task-level assessment of job satisfaction.Journal of Organisational Behaviour, 16, 2, 1995, pp. 101.

16.http://vpr.psych.umn.edu/msq.html

17. Source: Boundless. "Defining Job Satisfaction", Boundless Management. Boundless, 19 Jun. 2015. Retrieved 20 Jun. 2015 from

https://www.boundless.com/management/textbooks/boundless-management-textbook/organizational-behavior-5/drivers-of-behavior-44/defining-job-satisfaction-231-7247/

18. Iraj Mahdavi, Hamed Fazlollahtabar, Nezam Mahdavi-Amiri, Mohsen Arabmaghsudi, Norouz Kolbadinejad. Design of a Fuzzy Job Satisfaction Matrix with

Dynamic Performance Criteria. Proceedings of the 2011 International Conference on Industrial Engineering and Operations Management, Kuala Lumpur, Malaysia, January 22 - 24, 2011,pp. 1173-1178

19.www.systemdynamics.org/conferences/.../272.p... A First Approach to Model Satisfaction at Work Under Equity Theory Using Fuzzy Set Theory and System Dynamics.

20. Piegat A. Fuzzy Modeling and Control. In: Studies in fuzziness and soft computing. New York: Physica-Verlag,2000.

21. Aliev RA, Aliev RR. Soft Computing and its Application. World Scientific, New Jersey, London, Singapore, Hong Kong, 2001, - 444 p.

22. Aliev RA, Fazlollahi B, Aliev RR. Soft Computing and its Application in Business and Economics. Springer, 2004, pp. 446.

23. Aliev RA, Bonfig KW, Aliev FT. Messen, Steuern undRegeln mit FUZZY-LOGIK. Franzis-Verlag GmbH, Munchen, Germany, 1994, - 316 p. (In German)

24. Aliev RA. Fuzzy Expert Systems. In Aminzadeh F., Jamshidi M. (Eds). Soft Computing: Fuzzy Logic, Neural Networks and Distributed Artificial

Intelligence. NJ: PTR Prentice Hall,1994, 99-108p.