International Journal of Scientific & Engineering Research, Volume 5, Issue 1, January-2014 36
ISSN 2229-5518
Fuzzy Logic based Expert System for Students’ Performance Evaluation in Data Grid
Environment
Dr. Sanskruti Patel, Dr. Priti Sajja, Dr. Atul Patel
Abstract— In the current competitive era, every school, collages and institutes strongly focus on improvement of the students’ performance by applying fair evaluation methods. Performance of students is difficult to assess before the final results are declared. Major factors that affect on performance are the attendance of students in their theory and laboratory sessions, the exam grades they obtained in internal exams and the exam grades they obtained during term work evaluation. Assessing students’ performance is an area in which strict rules often do not represent the realistic situation. Fuzzy logic provides a way of representing the behavior of systems which are either too complex or too imprecise. Also, university is a very large domain and students’ data are generally stored in various distributed and heterogeneous data repositories which may span across different organizations and administrative domains. Such distributed data repositories can be integrated within data grid environment by implementing data grid middleware. Therefore, we have developed a fuzzy logic based expert system which assists the process of decision making of students’ performance evaluation within data grid environment. The system will utilize the fuzzy logic theory and develop the decision making process based on fuzzy rules to assess whether a student gets very poor, poor, good, average or excellent performance.
Index Terms— Fuzzy Logic, Expert Systems, Data Grid, OGSA-DAI, Agent, Multi-agent Systems, Fuzzy Logic
—————————— ——————————
IJSER
In the current competitive era, every school, colleges and in-
stitutes strongly focus on improvement of the students’ perfor-
mance by applying fair evaluation methods. There are several
factors are needed to be considered which affect the perfor-
mance. Major factors that affect on performance are the attend-
ance of students in their theory and laboratory sessions, the ex-
am grades they obtained in internal exams and the exam grades
they obtained during term work evaluation. It is possible to
analyze these factors qualitatively such as poor, very good, av-
erage, excellent etc. Even though quantitative figures are availa-
ble such as 55% attendance in theory session, 80% marks in in-
ternal exam, 70% marks in the term work evaluation; it cannot directly be used for performance assessment. Suppose in the grading process of students’ performance, the borderline be- tween a good performance and a bad performance is 50%. Sup-
pose the student got 49% during the grading process. Can this performance considered as a bad? Or if it is 50%, can it consider as a good? So, assessing students’ performance is the area in which strict rules often do not represent the real situation.
---------------------------------------------------------------------------------
Also, can two students have the same final exam grade, one
of which has been present in the majority of the sessions and the
other only in half of the sessions? These are the criteria accord-
ing to which the teachers can realistically assess students in or-
der to evaluate their performance. There are two main reasons
why above classical logic systems cannot deal with problems in
which knowledge is approximate. They do not provide a means
for representing the meaning of propositions expressed in a
natural language when it is imprecise, and they do not provide
a mechanism for inference in the cases where knowledge is rep-
resented symbolically along with its meaning [3].
As we can clearly see from the above discussion, developing a system for students’ performance evaluation based on strict and rigid rules would not be a truthful evaluation process. Therefore, we need systems that will deal with knowledge,
which is rather imprecise or incomplete as human routinely and subconsciously place things into classes whose meaning and significance are not well defined. Fuzzy logic provides a way of representing the behavior of systems which are either too com- plex or too imprecise.
Moreover, university is a very large domain and students’ data are generally stored in various distributed and heteroge- neous data repositories which may span across different organ- izations and administrative domains. Such data repositories can be integrated within data grid environment by implementing data grid middleware. Therefore, we have developed a fuzzy logic based expert system which assists the process of decision making of students’ performance evaluation within data grid environment. The system accomplishes decision making pro- cess to assess students’ performance. The implemented experi- mental system leverages of OGSA-DAI (Open Grid Services Architecture - Data Access and Integration) workflow execution functionality and out-of-the-box activities for data access, inte-
IJSER © 2014 http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 5, Issue 1, January-2014 37
ISSN 2229-5518
gration, transformation and delivery.
To establish a data grid environment, a data grid middle- ware technology named OGSA-DAI (Open Grid Services Archi- tecture – Data Access and Integration) has been used. The OG- SA-DAI has been studied and set of mechanisms have been identified to use data grid services. Open Grid Services Archi- tecture - Data Access and Integration (OGSADAI) is a project that develops middleware solution to assist with access and integration of data from separate data resources via the data grid [2].
The project was funded by the UK e-Science core program and is working closely with the Global Grid Forum Data Access
and Integration Services (GGF DAIS) work group [1]. OGSA- DAI is a database centric middleware solution which mainly provides the data virtualization services.
Configuration and deployment of data grid is a complex and lengthy task. Therefore, data grid services should be offered with minimum user intervention. To facilitate the access of data grid middleware services, we have implemented different col- laborative agents those are working together to provide the core, user & knowledge level services to the users.
Users are interacting with agents in order to access OGSA- DAI services. Agents are autonomous, adaptive and coopera- tive by nature and emerging as dynamic, flexible and extensible mediators for facilitating grid services in a data grid environ- ment [4, 6]. Agents are working proactively with minimum user intervention and offer various data grid services to its users [7].
IJSER
Figure 1: An Architectural View of a Fuzzy Logic based Expert System for Students’ Performance Evaluation in Data Grid Environment
agent system provides a scalable environment to integrate this
In a university, there are several different departments exist- ing and they may manage their students’ data individually and in disparate form. The system is able to integrate these distrib- uted and heterogeneous data repositories in the data grid envi- ronment. As discussed earlier, a data grid is realized through OGSA-DAI, a data grid middleware. As configuration of the data grid is a complex and tedious tasks, data grid formation should be performed with minimum human intervention. Agents are the sophisticated software entities that work proac- tively on behalf of its users. So, we have developed an agent- based interface to provide data grid services to users. The multi-
data grid to perform the knowledge extraction and delivery process and hides the heterogeneity of the database nodes from the users and handles the data access and integration in an effi- cient manner. Once the data grid is formed, we have integrated a knowledge-based component by realizing fuzzy logic. This system performs knowledge extraction and delivery of data or information retrieved from data grid which helps students and teachers to evaluate students’ performance and assist them in the decision making process at a certain level.
Figure 1 presents the architectural view of students’ perfor- mance evaluation system. We can see from the figure 1.1 that there are multiple collaborative agents working together as a part of multi-agent system environment. There are mainly three entities of the system. They are distributed & heterogeneous data repositories reside in data grid, data grid middleware and
IJSER © 2014 http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 5, Issue 1, January-2014 38
ISSN 2229-5518
multiple collaborative agents to configure and manage the data grid.
We can see the database nodes in figure 1 that contain rela- tional data resources (e.g. MySQL, Oracle, SQLServer etc.). These databases are located at different sites. Each site may con- tain different relational database type and schema. Also, each site contains necessary data to perform evaluation. i.e. attend- ance, exam marks and termwork evaluation marks of respective student. We have used horizontal fragmentation strategy where students’ databases may store at different sites (in different de- partments), having different data tuples but having similar table schemas. More departments (i.e. database nodes) can be added in the data grid according to requirement with help of agent- based interface. The agents access and integrate the necessary data (attendance, exam marks and termwork evaluation marks) from this database grid through OGSA-DAI and then apply fuzzy set theory and logic to provide knowledge-based assis- tance. Apart from relational resources (MySQL, Oracle, etc.), XML and file-based data resources may also integrate with nec- essary modification.
We have developed the following domain specific agents: User Management Agent, Student Profile Agent, Progress Mon-
The agent-enabled data grid model offered core, data and user level services. Through these services, the client applica- tions and users may get access to digital objects of the data grid environment. The typical agents namely resource management agent, data access & integration agent and grid administrative agent provide the above said services [7]. Table 1.2 shows the responsibilities of each data grid service agents.
itoring Agent, Path Advising Agent (A fuzzy Interface Agent), Search Agent and Report Generator Agent. The following is the detail of the services offered by each agent. Table 1.1 shows the responsibilities of each agent.
We have developed a fuzzy interface agent (i.e. path advising agent) which implements fuzzy inference process on behalf of users. There are mainly five components in a fuzzy inference pro- cess execution. They are Crisp Values, Fuzzification, Fuzzy Infer- ence Engine, Fuzzy Knowledge base and Defuzzification. The fuzzy expert system takes three input parameters for its calcula- tions: Attendance, Internal Exam Marks and Teamwork Marks and generates one output parameter as a result i.e. Performance. The input parameters are supplied by the data access & integra- tion agent. Data access & integration agent takes all these values from distributed and heterogeneous databases that reside in a data grid environment. The fuzzy inference agent matches the input provided by the corresponding rules of the rule base and produces a result fuzzy set. The given output defuzzifies value of the result fuzzy set. Students’ performance evaluation with a fuzzy expert system comprised with three steps:
IJSER © 2014 http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 5, Issue 1, January-2014 39
ISSN 2229-5518
Fuzzification of input parameters is carried out using input variables and their membership functions of fuzzy sets. We have formed a set of linguistic variables with respect to their input pa- rameters as well as output parameter. Fuzzification of perfor- mance is carried out by using three input parameters and their associated membership functions. For each input parameter, we have defined three linguistic values and membership functions. The fuzzy set of all three input parameters (attendance, internal exam evaluation and term work evaluation) and their associated membership functions are given in Table 1.3.
For output variable, we have defined five linguistic values and membership functions. The fuzzy set for output parameter (per- formance) and its associated membership functions are given in Table 1.4.
Linguistic Values | Membership Function | Interval |
Poor | Trapezoidal R | (3,5) |
Good | Triangular | (3,5,7) |
Excellent | Trapezoidal L | (5,7) |
RULE 10: IF Attendance IS good AND Exam_Evaluation IS poor AND Termwork_Evaluation IS poor THEN Performance IS VeryPoor;
RULE 11: IF Attendance IS good AND Exam_Evaluation IS poor AND Termwork_Evaluation IS good THEN Performance IS Average;
RULE 12: IF Attendance IS good AND Exam_Evaluation IS poor AND Termwork_Evaluation IS excellent THEN Performance IS Good;
RULE 13: IF Attendance IS good AND Exam_Evaluation IS good AND Termwork_Evaluation IS poor THEN Performance IS Average;
RULE 14: IF Attendance IS good AND Exam_Evaluation IS good AND Termwork_Evaluation IS good THEN Performance IS Good;
RULE 15: IF Attendance IS good AND Exam_Evaluation IS good AND Termwork_Evaluation IS excellent THEN Performance IS Good;
RULE 16: IF Attendance IS good AND Exam_Evaluation IS excellent AND
Termwork_Evaluation IS poor THEN Performance IS Good;
RULE 17: IF Attendance IS good AND Exam_Evaluation IS excellent AND Termwork_Evaluation IS good THEN Performance IS Good;
RULE 18: IF Attendance IS good AND ExamEvaluation IS excellent AND Termwork_Evaluation IS excellent THEN Performance IS Excellent;
RULE 19: IF Attendance IS excellent AND Exam_Evaluation IS poor AND Termwork_Evaluation IS poor THEN Performance IS Poor;
RULE 20: IF Attendance IS excellent AND Exam_Evaluation IS poor AND Termwork_Evaluation IS good THEN Performance IS Good;
RULE 21: IF Attendance IS excellent AND Exam_Evaluation IS poor AND Termwork_Evaluation IS excellent THEN Performance IS Good;
RULE 22: IF Attendance IS excellent AND Exam_Evaluation IS good AND
Termwork_Evaluation IS poor THEN Performance IS Good;
RULE 23: IF Attendance IS excellent AND Exam_Evaluation IS good AND Termwork_Evaluation IS good THEN Performance IS Good;
RULE 24: IF Attendance IS excellent AND ExamEvaluation IS good AND Termwork_Evaluation IS excellent THEN Performance IS Excellent;
RULE 25: IF Attendance IS excellent AND Exam_Evaluation IS excellent
Fuzzy rules are the rules determine input and output mem- bership functions that will be used in inference process. The deci- sion, which the fuzzy inference system makes, is derived from the rules. These rules are a set of ‘If-Then’ statements that are intui- tive and easy to understand as they are written in common Eng- lish declarations. We have derived and framed the following rules for the system to generate an expert advice on students’ performance evaluation.
RULE 1: IF Attendance IS poor AND Exam_Evaluation IS poor AND Termwork_Evaluation IS poor THEN Performance IS VeryPoor;
RULE 2: IF Attendance IS poor AND Exam_Evaluation IS poor AND Termwork_Evaluation IS good THEN Performance IS VeryPoor;
RULE 3: IF Attendance IS poor AND Exam_Evaluation IS poor AND Termwork_Evaluation IS excellent THEN Performance IS Poor;
RULE 4: IF Attendance IS poor AND Exam_Evaluation IS good AND Termwork_Evaluation IS poor THEN Performance IS VeryPoor;
RULE 5: IF Attendance IS poor AND Exam_Evaluation IS good AND
Termwork_Evaluation IS good THEN Performance IS Average;
RULE 6: IF Attendance IS poor AND Exam_Evaluation IS good AND Termwork_Evaluation IS excellent THEN Performance IS Good;
RULE 7: IF Attendance IS poor AND Exam_Evaluation IS excellent AND Termwork_Evaluation IS poor THEN Performance IS Poor;
RULE 8: IF Attendance IS poor AND Exam_Evaluation IS excellent AND Termwork_Evaluation IS good THEN Performance IS Good;
RULE 9: IF Attendance IS poor AND Exam_Evaluation IS excellent AND
Termwork_Evaluation IS excellent THEN Performance IS Good;
AND Termwork_Evaluation IS poor THEN Performance IS Good;
RULE 26: IF Attendance IS excellent AND ExamEvaluation IS excellent
AND Termwork_Evaluation IS good THEN Performance IS Excellent;
RULE 27: IF Attendance IS excellent AND ExamEvaluation IS excellent
AND TermworkEvaluation IS excellent THEN Performance IS Excellent;
Inference Method: An output fuzzy set is obtained from the input variables and rules by applying the inference procedure. We have used the Mamdami’s max-min inference method as it is typically used in modeling human expert knowledge and it is given below [5]:
Defuzzification is the process of converting a fuzzy number in- to a crisp value. It is the process of producing a quantifiable result in fuzzy logic based on given fuzzy sets and corresponding membership degrees. Several methods have been developed so far for defuzzification. In our experimental system, we have used Center of Gravity (CoG) Method (also known as CoA – Center of Area) for defuzzification.
It is observed that, to store the data of students, the different institutes may use different storage systems and data formats
IJSER © 2014 http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 5, Issue 1, January-2014 40
ISSN 2229-5518
with differences in their schema, access rights, metadata attrib- utes and many more. Therefore, there is a need of the system which provides a single and a virtualized view of such institu- tional data assets and allows data provides of institutes to retain control of their data. As a result, we have developed a system which operates in a data grid environment and new institution may add as a new database grid node in the system without much modification in code. Also, the system offers scalability and extensibility and implemented a strategy used to avoid a single- point-of-failure. The system evaluates the performance of the students at regular interval, generates the appropriate reports for the same and makes these reports available to the students. The reports are generated on regular interval as a part of a continuous evaluation process and provide a chance for the students to in- crease their performance rate before the final exams are conduct- ed. The university may accomplish the assessment of their stu- dents’ performance in a timely and user friendly manner. Also, the performance may disaggregate by institutes, instructor, gen- der, ethnicity, economic status and disability; have a meaningful impact on the achievement of students throughout their career.
This research work presented here enhances business intelli- gence and quickening decision making process by providing knowledge-based component through fuzzy set theory and fuzzy logic that leads to a novel model. A fuzzy logic based expert sys- tem evaluates students’ performance on regular basis and advise them on their state of progress. The system uses the data grid service agents in order to expose the heterogeneous and distrib- uted databases in grid environment. Moreover, it implements the fuzzy interface agent to realize the knowledge-based component, which assists the users in the decision making process at a certain level.
[1] Antonioletti M, Krause A, Paton NW, An outline of the Global Grid forum data access and integration service specifications. Data Man- agement in Grids, 3836, doi: 10.1007/11611950_7, Pages: 71-84, 2005
[2] Karasavvas K, Antonioletti M, Atkinson M, Hong NC, Sugden T,
Hume A, Jackson M, Krause A, Palansuriya C., Introduction to OG-
SA-DAI Services, Volume 3458, Pages:1-12, 2005
[3] L.A. Zadeh, Fuzzy logic, Computer, Volume 21, Issue 4, Pages: 83-93,
1988.
[4] N. R. Jennings, M. Wooldridge, Applications of Intelligent Agents,
retrieved from http://www.cs.ox.ac.uk/people/michael.wooldridge/pubs/applica tions.pdf on February 2013
[5] Ramjeet Singh Yadav, Vijendra Pratap Singh, Modeling Academic Performance Evaluation Using Soft Computing Techniques: A Fuzzy Logic Approach, International Journal on Computer Science and En- gineering (IJCSE), Volume 3, No. 2, Pages: 676-686, 2011
[6] S.genesereth M., Ketchpe, Software agents, Comm. of ACM, Volume
37, No. 7, Pages : 48-53, 1994
[7] Sanskruti Patel, Priti Sajja A Grid-based Model for Integration of
Geographically Distributed & Heterogeneous Educational Resources for Knowledge Extraction & Delivery, , International Journal of In- novative Research in Science, Engineering and Technology, Vol. 2, Is- sue 9, September 2013, pp.: 4612-4617
IJSER © 2014 http://www.ijser.org