Proceedings oI International ConIerence on Computing Sciences
WILKES100 ICCS 2013
ISBN: 978-93-5107-172-3 Fuzzy rule based expert system Ior employee appraisal based on UGC guidelines Nidhi Gumber 1* , Prateek Aggarwal 1 , Sanjay Kumar Singh 1 and Leena Jain 2
1 Department of Computer Science and Engineering, Lovely Professional University, PB, India 2 Department of Computer Applications, Global Institute of Management and Emerging Technologies, Amritsar, PB, India Abstract Employee Appraisal system is an important strategy oI any organization to manage the human resources. A lot oI Iactors are important Ior an employee to evaluate and improve their perIormance. Employee`s promotions, salary bonus and many other incentives are based only on their perIormance in organization. In this paper, a Iuzzy-logic approach is implemented to calculate the employee`s perIormance at higher university level considering the Iactors deIined by UGC guidelines that have major eIIect on an employee`s perIormance. A Iuzzy rule-base is implemented using the critical Iactors. And then all the deIuzziIication techniques are applied on the data oI employees to rank those employees. 2013 Elsevier Science. All rights reserved. Keywords: Fuzzy Expert System, API Scores, Fuzzy InIerence System, FuzziIication, DeIuzziIication 1. Introduction PerIormance appraisal is a Iormal management system that provides Ior the evaluation oI the quality oI an individual`s perIormance in an organization. PerIormance appraisal is the procedure to evaluate the current and past perIormance oI an employee. (Moon, 2007) There are various techniques and Iactors used to evaluate the perIormance oI en employee. As outlined by GMeenakshi, various Iactors include Planning and Preparation oI course (eIIectiveness and punctuality), Maintaining classroom discipline and control on class, Communication and knowledge skills and ProIessional responsibilities and contributions towards college and society. On other hand, Abdur Rashid Khan includes some more Iactors like Research contribution to the evaluate employee appraisal. Fuzzy set theory is very common Ior developing an expert system. Chiang and Lin have proposed a method Ior teaching assessment using Iuzzy system (T.T.Chiang, 1994). Chen and Lee have proposed two diIIerent methods to evaluate the students using Iuzzy sets (S.M.Chen, 1999). Similarly there exist so many expert systems which use Iuzzy set theory. Various techniques are present Ior evaluating employees like Rating, Trait scales, criteria based, management by objectives (MBO), 360 degree Ieedback, work planning and review and peer review. (C. C. Yee, 2010) All techniques have their own advantages and disadvantages, most organizations use all oI these techniques together to work eIIectively. 2. Design of Fuzzy Expert System Design oI Iuzzy expert system is shown in Iig 1.1. This Iigure describes the process, how to sequentially create this Iuzzy expert system. These steps are discusses one by one below. * Corresponding author - Nidhi Gumber 80 Elsevier Publications, 2013 Nidhi Gumber, Prateek Aggarwal, Sanfav Kumar Singh and Leena Jain
2.1. Identification Of Factors This step is the initial step in our work where we need to identiIy the Iactors that will directly have an impact on the employee`s perIormance. These Iactors are mainly divided in three major categories, which we will discuss here: 2.1.1 Teaching, Learning and Evaluation related activities: This major Iactor is mainly considering the teaching related activities which are essential Ior a Iaculty member to perIorm. This major Iactor consists oI the Iollowing points: - 2.1.1.1 Essential: This category consists oI Iactors like Lectures taken as percentage oI lectures allocated, Tutorials hours undertaken as percentage oI actual allocated, Practical or Lab hours taken as percentage oI actual allocated, Imparting oI knowledge/ instruction with prescribed material, College/University examination duties (question paper setting, vetting, and evaluation), CA evaluation duties, Additional examination duties like Ilying squad, invigilation etc. 2.1.1.2 Desirable: Desirable is another sub-category oI Iirst major category. This category consists oI some points which depend on the desire oI a Iaculty member, whether he/she wants to perIorm these tasks or not. This includes Iactors like Instruction Plan designing, Scheme creation, Syllabus catalogue designing, Course catalogue designing, Conduction oI any training or workshop etc. 2.1.2 Co-Curricular, Extension and Profession-related activities: This is another major category which involves the contribution oI a Iaculty member in co-curricular activities like sports, cultural committees, training and placement committee etc. This category again contains subcategories, which are described below: 2.1.2.1 Co-Curricular and Extension activities: This category involves Iactors like Contribution to Corporate liIe (Research coordinator, Academic Operation Coordinator, Master course coordinator and course coordinator), Institutional Co-Curricular activities Ior students (Placement / Training coordinator), Community Service and other student mentoring activities through NSS/NCC/NSO, Students and staII-related socio cultural and sports programs (Sports/Cultural committees), ProIessional ethics and campus development activities etc. 2.1.2.2 Profession-related contribution: This category consists oI the points that are concerned with the contribution given to the activities related with their proIession like Institutional Governance responsibilities, Membership in proIession related committees at state and national level, Participation in conIerences, seminars, Participation in Faculty Development Program, ProIessional Learning Committee, short term training courses less than one week, Membership oI proIessional associations committees, Boards oI Studies etc. 2.1.3 Research Publications: This is third major category, which is concerned Ior their contribution in research area. It involves research publications, text books written, journals published etc. It consists oI Iollowing sub-categories: - 2.1.3.1 Research Papers published: This category considers the papers published in journals with a good impact Iactor. This category includes Iactors like ReIereed journals with impact Iactor listed in Thomson Reuters, ReIereed journals not listed in Thomson Reuters but recognized journal, ConIerence proceedings as Iull paper in recognized conIerence etc. 2.1.3.2 Conference/ Seminars/ Workshop Papers: This category includes the conIerences, seminars or workshops that a Iaculty member has attended and presented any paper. The main Iactors included in this category are Papers in ConIerence/ Seminar/ Workshops (International ConIerence, National ConIerence), Invited Lectures or presentations (International level, National level) etc. 2.1.3.3 Student Mentorship for research guidance: This category will include the number oI students one Iaculty member has Ior research guidance. We consider the Iactors like Number oI M.Tech students Ior research guidance, Number oI M.Phil students Ior research guidance, Number oI PhD students Ior research guidance, Number oI Capstone project groups oI under-graduates Ior guidance etc. 2.1.3.4 Research Publications (Books, Chapters in Book, other than refereed journal): This category includes Iactors like Books published by International publisher, Books published by National level or State level publisher, Chapters contributed to the volumes published by International publisher, Chapters contributed to the volumes published by National publisher etc. 2.1.3.5 Research Projects: This category considers the projects that a Iaculty member is doing or has completed. The Iactors considered in this category are Sponsored Projects with budget( 50 lakh or above, 30 to 81 Elsevier Publications, 2013 Fuzzy rule based expert system for employee appraisal based on UGC guidelines
50 lakh, 10 to 30 lakh, 5 to 10 lakh), Completed Projects with budget( Above 30 lakh, Below 30 lakh, International) etc. 2.2. Fuzzification FuzziIication is a process oI converting crisp values oI an input to Iuzzy values. We do this by simply recognizing that many oI the quantities that we consider to be crisp and deterministic are actually not deterministic at all, they carry considerable uncertainty. (Ross, 2005) Table 1 Critical Factors Fuzzy linguistic input variables with membership range Fuzzy linguistic output variables with membership range Essential Iactors Poor Avera ge Good PerIormance in terms oI Essential Factors (0-100) poor Below avg avg Above avg Good Lectures, Labs, Tutorials (0-50) 0-20 10-30 20-50 0-25 20-35 30-45 40-60 55-75 Imparting oI knowledge (0-20) 0-8 5-15 12-20 Examinatio n Duties (0- 10) 0-3 1-7 5-10 CA evaluation( 0-10) 0-3 1-7 5-10 Additional duties(0-10) 0-3 1-7 5-10 Desirabl e Iactors Poor Avera ge Good PerIormance in terms oI Desirable Factors (0-50) poor Below avg avg Above avg good IP designing (0-10) 0-5 2-8 5-10 0-12.5 10-18 15-23 20-30 27-37 Scheme Creation (0-10) 0-5 2-8 5-10 Course Catalogue Design (0- 10) 0-5 2-8 5-10 Syllabus Catalogue design(0- 10) 0-5 2-8 5-10 Conductio n oI any training (0-10) 0-5 2-8 5-10 82 Elsevier Publications, 2013
The conversion can be done using membership Iunctions. Following are the tables prepared with expert views Ior the membership Iunction ranges and overlapping region. Table 1.1 is describing the subcategories oI Iirst main category oI Teaching, Learning and Evaluation related activities. It contains the linguistic variables as Low, Average and SatisIactory. The various ranges described here, are the regions deIined Ior these linguistic variables. Tables Ior all other categories are also prepared in the same way. Following Iigures are showing the Fuzzy InterIace in Fuzzy toolbox in MATLAB: - Fig 1.2 shows the sample IuzziIication Ior Essential Iactors, with ranges as deIined above. This sample contains Iive input values, each deIined with three membership Iunctions, describing Low, Average, and SatisIactory. Fig1.2: - Fuzzy Ior Essential Iactors Fig 1.3 shows the sample IuzziIication oI Desirable Iactors in Fuzzy Toolbox oI MATLAB. 1.3 Fuzzy Rule Formation Fuzzy rule base is prepared with IF-THEN rules, aIter Ietching knowledge Irom domain experts. These rules decide the output oI a crisp value given as input to Iuzzy. The input to an iI-then rule is the current value Ior the input variable and the output is an entire Iuzzy set. (Rohan) Here, as we are Iollowing a nested Iuzzy approach, we have designed diIIerent Iuzzy systems Ior diIIerent categories as well as sub-categories, and these Iuzzy systems have diIIerent number oI rules designed. Rules Ior Iuzzy inIerences are as shown in Iollowing Iigures. 83 Elsevier Publications, 2013 Nidhi Gumber, Prateek Aggarwal, Sanfav Kumar Singh and Leena Jain Fuzzy rule based expert system for employee appraisal based on UGC guidelines
Rule Base designed Ior Essential Category contains 243 rules, which are in IF-THEN Iorm. Sample oI these rules are as shown: - II (LecturesSeminarsTutorials is low) and (ImpartingoIknowledge is low) and (Examinationduties is low) and (CAandMteevaluation is low) and (InvigilationFlyingSquadduties is low) then (EssentialPerIormance is low). In similar way, all other Iuzzy systems are designed. Fig 1.4 shows the rule base designed Ior all categories including 1. FirstCat (Teaching, learning and evaluation related activities) 2. SecondCat (Co-Curricular and proIession-related activities) 3. ThirdCat (Research and academic contribution). This Iuzzy will calculate the Iinal scores which a Iaculty member will achieve considering all the Iactors. In this Iuzzy, there are three inputs and one output, with 27 rules designed. Some sample rules designed in Iorm oI IF-THEN are shown below: - II (FirstCat is poor) and (SecondCat is average) and (ThirdCat is poor) then (Iinalscores is belowavg). II (FirstCat is poor) and (SecondCat is average) and (ThirdCat is good) then (Iinalscores is average). II (FirstCat is average) and (SecondCat is belowavg) and (ThirdCat is belowavg) then (Iinalscores is average). II (FirstCat is aboveavg) and (SecondCat is good) and (ThirdCat is good) then (Iinalscores is good). II (FirstCat is excellent) and (SecondCat is good) and (ThirdCat is average) then (Iinalscores is good). Fig 1.4: - Rule base Ior all categories 2.3. Inference System Design Mamdani system is used in our system to design the inIerence, as it is most popular inIerence system to be used. In Mamdani inIerence system, Max-Min composition is used. Mamdani system Iollows 4 steps to calculate the Max-Min composition oI each rule. The steps are: - Evaluate the antecedent Ior each rule. Obtain each rule`s conclusion. Aggregate conclusions oI all rules. DeIuzziIication. Given the inputs as crisp values, it gets the membership values. II the rule have two antecedent parts and both are connected by OR operator (disjunction), then it will calculate the result oI rule by applying OR operation (max value). And iI both antecedent parts oI rule are connected by AND (conjunction) operator, then it will calculate the result oI rule by applying AND operation (min value). Now when we get the antecedent value oI rules Irom step 1 and as we have consequent part oI the rule, it applies Iuzzy implication method on it, which will truncate the membership value oI the consequent. 84 Elsevier Publications, 2013
AIter getting truncated values oI each rule, it aggregates the values oI each rule to get a single Iuzzy set. And then by applying deIuzziIication, it gets a single value representing the whole set. (Alonso, 2013) The overall inIerence system design is as shown in Iigure 1.5 below: - Fig 1.5: - InIerence system design 1.5 Defuzzification In simple words, we can say deIuzziIication is to Translate results back to the real world values. DeIuzziIication is the reverse process oI IuzziIication. The task oI DeIuzziIication is to Iind one single crisp value that summarizes the Iuzzy set that enters it Irom the inIerence block. (Lyons, 2012) The most popular deIuzziIication method is the centroid calculation, which returns the center oI area under the curve. There are Iive built-in methods supported: centroid, bisector, middle oI maximum (the average oI the maximum value oI the output set), largest oI maximum, and smallest oI maximum. We have applied all the methods on input sets and compare their ranks. 1.6 Evaluating and Testing the Results For evaluating and testing the results, we are using the employee`s data oI Lovely ProIessional University, Phagwara. The inIerence engine used here is Mamdani Iuzzy system. Results oI these employees are as shown in table 1.2. 85 Elsevier Publications, 2013 Nidhi Gumber, Prateek Aggarwal, Sanfav Kumar Singh and Leena Jain Fuzzy rule based expert system for employee appraisal based on UGC guidelines
Table 1.2: - Results aIter evaluation F a c u l t y
I D F i r s t
C a t e g o r y
( 0 - 7 5 ) S e c o n d
C a t e g o r y
( 0 - 3 0 ) T h i r d
C a t e g o r y
( 0 - 4 5 )
DeIuzziIied Output and ranks Centroid ranks Bisector ranks MOM ranks LOM ranks SOM ranks F1 48.33 8.1435 13.2025 42.7 3 42.8 3 42.8 4 42.8 4 42.8 4 F2 45.789 3.5604 13.2025 42.7 3 42.8 3 42.8 4 44.3 4 41.3 5 F3 48.33 3.5544 13.2025 42.7 3 42.8 3 42.8 4 42.8 4 42.8 4 F4 37.498 3.5604 13.2025 42.7 3 42.8 3 42.8 4 42.8 4 42.8 4 F5 29.66 3.5575 13.2025 32.7 4 32.2 4 33 5 33.7 5 32.2 7 F6 28.656 3.5575 13.2025 32.7 4 32.2 4 32.2 6 32.2 6 32.2 7 F7 37.4 3.5604 31.7354 52.7 2 53.4 2 52.6 3 53.4 2 51.9 3 F8 42.981 3.5574 13.2025 42.7 3 42.8 3 42.8 4 45.8 3 38.3 6 F9 37.49 11.216 13.2025 42.7 3 42.8 3 42.8 4 42.8 4 42.8 4 F10 63.53 3.5604 13.2007 52.7 2 53.4 2 53.4 2 53.4 2 53.4 2 F11 48.33 11.218 31.7224 67.4 1 67 1 67 1 68.5 1 65.4 1 F12 48.33 11.218 22.4364 52.7 2 53.4 2 53.4 2 53.4 2 53.4 2 3. Conclusion This paper proposes a method Ior evaluating the perIormance oI employees based on various Iactors as considered by UGC. This system is implemented on higher university level. It can also be implemented in any university recognized by UGC. The output oI system is entirely dependent on the API scores that a Iaculty obtained Irom his activities, more the contribution he will make more scores he will get. In Iuture, this system can be extended to the neuro-Iuzzy approach, which will help it to improve its results more. References |1| Albar, F.M. (2008). Employee Rating Systems and Innovations. PICMET Proceedings, 1-6. |2| Alonso, S.K. (2013). Mamdani`s Fuzzy inIerence method. Retrieved 2 28,2013, Irom dma.Ii.upm.es: http://www.dma.Ii.upm.es/java/Iuzzy/IuzzyinI/mamdanien.htm |3| C.C Yee, a.Y. (2010). PerIormance Appraisal System using MultiIactorial Evaluation Model. International Journal oI Human and Social Sciences, 1-5. |4| Fuzzy Membership Functions. (n.d.). Retrieved Irom tech.dmu.ac.uk: http://www.tech.dmu.ac.uk/~hseker/CSCI340620200809/Week20-2032020CSCI20300620Lecture203and4.ppt |5| Haykin, S. (1998). Neural Networks, A comprehensive Ioundation. Prentice Hall. |6| Knapp,B.(2013).Fuzzy set and pattern recognition. Retrieved Irom cs.princeton.edu: http://www.cs.princeton.edu/courses/archive/Iall07/cos436/HIDDEN/Knapp/Iuzzy002.htm |7| Lyons, G. (2012). DeIuzziIication. Retrieved 2 24, 2013, Irom ecestudents.ul.ie: www.ecestudents.ul.ie/CoursePages/MSc.../Iuzzy203.pdI |8| MathWorks.in.(n,d.). Fuzzy InIerence Process. Retrieved Irom mathworks.in: http://www.mathworks.in/help/Iuzzy/Iuzzy-inIerence- process.html |9| Moon, C.L (2007). An Implementation case Ior the PerIormance Appraisal and Promotion Ranking. IEEE International ConIerence on System, Man and Cybernetics. 86 Elsevier Publications, 2013
|10| Moore, J.S.(1990). An Expert System Prototype Ior PerIormance Appraisal oI Managerial Communication. Indiana University at Fort Wayne, 1-12. |11| Robert S. Engelmore, E.F. (1993, May). Expert Systems and ArtiIicial Intelligence. Retrieved Irom www.wtec.org: http://www.wtec.org/loyola/kb/c1s1.htm |12| Rohan. (n.d.). Fuzzy Logic Toolbox. Retrieved Irom www.rohan.sdsu.edu: http://www- rohan.sdsu.edu/doc/matlab/toolbox/Iuzzy/Iuzzytu5.html |13| Ross, T. (2005). FuzziIication. T.Ross, Fuzzy logic with engineering applications. John Wiley & Sons. |14| S.M.Chen, C. (1999). New methods Ior student`s evaluating using Iuzzy sets. Fuzzy Sets and Systems, 209-218. [15] T.T.Chiang, C. (1994). Application oI Iuzzy theory to teaching assessment. National ConIerence on Iuzzy theory and applications, Taipei, Taiwan, 92-97 87 Elsevier Publications, 2013 Nidhi Gumber, Prateek Aggarwal, Sanfav Kumar Singh and Leena Jain Index
A API scores, 86
D Defuzzification, 85
F FES. see Fuzzy expert system (FES) Fuzzification process, 8283 Fuzzy expert system (FES), 80 defuzzification, 85 evaluating and testing, 8586 factors, identification of, 8182 for essential factors, 83 fuzzification process, 8283 inference system design, 8485 rule formation, 8384 toolbox of MATLAB, 83 Fuzzy rule formation, 8384