Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Students’ Performance and Employability Prediction through Data Mining: A Survey

Students’ Performance and Employability Prediction through Data Mining: A Survey Objective: To systematically review the work done in the field of academic performance prediction and employability prediction of students in higher education. Methods: The survey first explain show higher education has become an exciting field of research and why the prediction of academic performance and employability is beneficial for the institutions. We also explain briefly in how many ways higher education is being provided world-wide. Then we discuss the work done in both the areas of prediction. Findings: The survey explores existing research highlights and finds that prediction of academic performance has progressed a lot but employability prediction is yet to mature. Application: It further suggests few parameters that have not been considered so far in predicting the performance or employability. Keywords: Academic Performance, Data Mining, Employability, Higher Education, Prediction, Survey e r Th eview is organized chronologically and 1. Introduction categorically to oer in ff sight on how past research efforts laid the groundwork for subsequent studies, including In the present knowledge-based epoch, education plays the present research efforts. The detailed review has been a major role in the progress of a nation’s economy and carried out, so that; the present research can properly be development. It assures to invigorate the country by tailored to add the present body of literature, as well as the contributing the reliable and quality workforce to the scope and direction of the present research effort. society. Higher education is the foundation for fostering In order to give a comprehensive view of the work the talent, the key factor in increasing national human done in the field of EDM, many survey papers have been capital quality, and the main way to upgrade a nation’s published. The most cited research compiled the work competitive status. Thus, the research on development done till 2005 and the other that has discussed vital of higher education is an important work and is actually features of EDM . The work done in EDM till 2010, 2013, required. To get an edge over each other, institutions are 3–5 2014 has been complied in the different research papers applying cutting edge technologies like data mining on (Table 1). the huge data generated in class room including academic, According to the literature study, EDM research behavioral, demographic data of students and faculty pertains mainly three heads, according to the way data data as well. e Th data generated in educational set up can is collected- give deep insight into educational process. Educational • Traditional face to face or the offline education Data Mining (EDM) refers mining the data generated in system based on data generated in the classroom. educational set up. • E-learning in which the learning is provided through Researches in EDM have benefited the educational online content based on online activity logs. setups tremendously. * Author for correspondence Students’ Performance and Employability Prediction through Data Mining: A Survey • Intelligent tutoring system (ITS) and Adaptive 3. Prediction of Students’ Educational Hypermedia System (AEHS) involve Performance online teaching based on students need, his or her progress rather than providing same structured Students’ academic performance is a mature field now lesson to all the students. with many researchers contributing to it. Moving in chronological increasing order from 2007 onwards the Table 1. Survey Papers in Educational Data Mining researchers have considered various parameters for References Highlights prediction and using different algorithm they have tried Comprehensive survey for traditional to predict the result of specific course of a particular educational systems. university. Highlights the work done in EDM from the Authors have considered students from two different view point of models, emergence of public data, tools. institutions, one International and one small institution, Illustrates problems of education system to check the effect of research separately on both. In both resolved using data mining and proposed the the cases decision tree provided better accuracy than association of techniques for better predictions. Bayesian Network. Recognized different educational system tasks, A simple student performance assessment and discipline, techniques, and algorithms. monitoring system based on various data mining Provides a comprehensive survey, a travelogue techniques was developed with the predictor attributes from 2002 to 2014 for educational data mining. including students’ demographic details, course average st th 7 score in 1 to 5 semester overall gain performance, etc . E learning and Intelligent Tutoring systems have used Decision Tree C5 showed highest accuracy, followed by EDM mainly to model online behavior of students, track Classification and Regression, Trees (CART), Artificial their performance and get feedback from them. Neural Network (ANN), Chi-Squared Automatic However, this survey explores traditional educational Interaction Detection (CHAID). settings as it is still most widely used method of teaching Taking more factors like university matriculation across the globe. exam, GCE (General Certificate of Education) Score, Senior Secondary Certified Examination score SSCE), grades in O level subject, location of University from 2. Traditional Educational Setup home, gender and age, Cumulative Grade point was to Traditional educational set up refers to class room teaching classified as Good, Average and Poor. Artificial Neural which is still the most popular form of teaching by the Network (ANN) was used for prediction and 74.5 Institutions across the globe and needs to be explored. %accuracy was attained in performance prediction . Students prefer to take admission in an Institution with Prediction of school students’ performance has high academic performance and where the passing out been considered with a total of 33 parameters including students have better employability. Prediction of both socio -demographic details like (parental marital status, academic performance and employability can help the father’s job, mother’s job, quality of family relationship, management identify students at risk of poor academic attitude towards study (No. of hour, past failure) Internet performance and low employability. The process of facility, family support, free time aer s ft chool, health, prediction involves application of various data mining alcohol consumption etc . To predict the performance, algorithms, to predict the dependent variables based on Decision was used. Attendance, parents job, previous year independent factors. performance was found to be the key factors that ae ff ct Next two subsections discuss the researcher’s the current achievement. inclination towards prediction of both students’ e Th study uses feature selection process of Waikato performance and their employability. The focus is on Environment for Knowledge Analysis (WEKA) tool the attributes considered and methods/classification which has inbuilt set of methods and considers Student’s algorithms adopted for prediction of performance and gender, Eyesight, Community, Physical handicap, food employability prediction. habit, family details, mode of transport, medium of Vol 10 (24) | June 2017 | www.indjst.org Indian Journal of Science and Technology Tripti Mishra, Dharminder Kumar and Sangeeta Gupta instruction, sports activity etc. as predictive factors.It consider demographical details of the students, but was also observed that classification methods like Naïve concentrate only on grades of various courses as attributes Bayes, one R voted perception performed much better to predict categories of students as first class, the other with feature selected subset than where all variables were consisting second class upper lower and third class. In this considered. case Naive Bayes was found to show better accuracy than Voting technique is a method of applying more than Decision Tree. In a contrast authors have used previous one algorithm in succession and then taking the best academic achievements in exams and present academic result. e Th researchers have concluded that Decision assessment in Lab and class assignments to predict end Stump along with Hidden Naïve Bayes is most suited for semester marks, whereas another research paper has academic performance prediction in case of a New Zealand made academic performance prediction of Engineering Polytechnic, where, apart from academic performance, Drawing course. SVM is well suited for individual demography, disability was also considered as predictive students’ performance prediction and regression is to be factors. In a similar research CART algorithm has used for prediction of entire class. 13 20 been found to give highest accuracy. Further, the study In another study language proficiency in English, added few more predictive parameters like ethnicity, and credit selection and whether a student is unmarried, student’s current job condition to predict performance. In married, divorced have shown their effect on academic this case tree was found less accurate than regression and performance prediction and decision tree showed better analysis. accuracy over neural network. Chi Squared Automatic Interaction Detector eTh research used various factors that were (CHAID) was used make high school result prediction ranked according to their effectiveness in predicting with predicting factors being taken as health of the the placement result of Turkish secondary school. The student, tuition availability, facilities to study at home etc. predictors included marks Spirituality and morals, Prediction accuracy was not very good 44.69% and key Turkish language and entry level exam. Decision tree influences were found to be mother’s education, location worked well with the problem. from school etc . Demographic variable, score of high school entrance The study aimed to discover individual characteristics exam and attributes related to their attitude towards that decide their success using Microsoft Decision studying etc. were considered for predicting the result Tree.11 attributes that includes registered information of 1st year students of economics course . A set of four high school information; Turkish, University entrance algorithms were applied. It was found that past academic exams degree and University placement info family living records, entrance exam marks, hours put into studies are conditions and financial status etc. were considered. having maximum impact on prediction accuracy. Microsoft Decision Tree (MDT) has been used to predict Students’ academic performance in five categories GPA with just two categories successful and unsuccessful. were predicted using Decisions Tree, Functions, Rules, Four classification methods Artificial Neural Network, Bayes Net etc. and Random tree provided the best Decision trees, Support Vector Machine and logistic result. In a more recent study , factors considered were regression along with ensemble techniques (i.e. bagging, demographic profile, previous academic scores, entrance busting and information fusion) to analyze 16,000 exam result and among the various algorithms applied j48 students records . 39 variables including demographic gave highest accuracy of prediction. data, TOEFL, SAT Score, Loan etc. were used to predict In yet another research paper at school level attrition/ retention. Support Vector machine provided performance was predicted using Parental status, Mode Highest Accuracy followed by decision tree, Artificial of transport, Groups of subjects, type of school, previous Neural network and regression. marks and applying algorithms decision tree, KNN, SVM In another research paper attribute importance was etc. emphasized by ranking them, using correlation based Academic performance is always one of the primary feature subset selection and consistency subset selection predictors of employability. Thus all the factors ae ff cting (COE) and using them further find accuracy of various academic performance also ae ff ct the employability. Next classifiers . Unlike other researchers, authors do not we explore the work done on employability. Vol 10 (24) | June 2017 | www.indjst.org Indian Journal of Science and Technology Students’ Performance and Employability Prediction through Data Mining: A Survey were stronger predictor of their employability than 4. Prediction of Students’ their technical education consisting of their academic Employability performance . Poor English language competency is found to be a major reason for the low employability . Today, the reputation of an Institution is judged by This has been supported by one more study which its academic success, its ability to retain students concludes that knowledge of GPA and English Language and to provide employment for its students. The competency are required for the students in software term “Employability” still has no precise definition. industry to continue with their employment and also Employability has been described in many ways, like, the the female candidates were better performer than male ability to secure a job, getting a job within a specified time candidates in campus placement drives. period aer g ft raduating, the ability to skill map oneself Psychology seeks an answer to the kind of skill according to the job need, or the willingness of the student requirement which is essential for enhancing job to extend the graduate learning at work . Alternatively, prospects. A sound effect of overall personality of the employability is defined as the ability of students to employee along with career competence, self-efficacy is secure a job during on campus placements . Research seen on employability. This has been supported by the in employability prediction is in nascent stage. It mostly study where employers give priority to soft skills, team involves identification of skills or attributes required work, etc. from the perspective of employers and is obtained from Another researcher concludes that J48 is most employer through questionnaire and interviews. Mostly suited algorithm to predict the employability based statistical methods have been applied and research is on demographic profile, employees job satisfaction, more of descriptive than predictive. academics etc. e im Th portance of psychological factors along It is comprehensible that not much work has been with personal and organizational awareness has been done in the direction of employability prediction emphasized in a report by Higher Education Academy mainly due to lack of authentic data, hence this study of with the Council for Industry and Higher Education employability prediction and model development will (CIHE) in United Kingdom another study considered contribute significantly in educational data mining. the effect of working environment on performance of the Graduate employability has been the subject of little employee . empirical research. er Th e are a number of difficulties Whether an employee will be able to meet the in defining and measuring graduate employability, expectation and should be hired is predicted in research which means that there is a paucity of research that paper which is based on few attributes extracted from looks at its predictors and outcomes. Previous work has candidate’s curriculum vital key words in application and proposed that emotional competence improves graduate interview. employability. Researchers in psychology have shown It has been concluded that as the paradigm is shifting that the emotional skills of a student are also important from product based to service based industry especially factors for performance prediction. However not much in Information Technology, and hence the curriculum work has been done in the field of EDM to validate the and method of delivering lecture must evolve in order to existing knowledge or construct new knowledge about enhance employability . emotional skill being predictor of performance. One A descriptive study in the research indicates that reason for not considering emotional skill is lack of employers look forward to employees with Personal authentic data. The need of hour is to construct authentic attributes that include loyalty, commitment, honesty, primary data that has factors of academic integration integrity, enthusiasm, reliability, personal presentation, and social integration and emotional skills and study the common sense, positive self-esteem, and a sense of humor, prediction of academic performance and employability motivation, adaptability, a balanced attitude towards work in tandem. e Th authors of this paper have identified this and home life and ability to deal with pressure. gap and published two papers in this regard where apart Correlation and ordinal regression has been used from social and academic integration, emotional skills to conclude that a students’ non-technical education like leadership, self-esteem, empathy; decision making consisting of reasoning, logical ability, and soft skills Vol 10 (24) | June 2017 | www.indjst.org Indian Journal of Science and Technology Tripti Mishra, Dharminder Kumar and Sangeeta Gupta 2009: A review and future visions. JEDM-Journal of Educa- capability, time management and stress management tional Data Mining. 2009 Oct; 1(1):3–17 are also considered as predictor variables. A model 3. Romero C, Ventura S. Educational data mining: A survey was derived for performance prediction based on from 1995 to 2005. Expert systems with applications. 2007 academic, social and above emotional skills. Similarly, Jul; 33(1):135–46. Crossref Employability prediction model has been developed by 4. Pe-a-Ayala A. Educational data mining: A survey and a data the authors using authentic primary data that involves mining-based analysis of recent works. Expert systems with applications. 2014 Mar; 41(4):1432–62. Crossref social, academic as well as emotional skills. Future work 5. a Th kar P. Performance Analysis and Prediction in Educa- requires developing tools for prediction of performance tional Data Mining: A Research Travelogue. International and employability. Journal of Computer Applications. 2015 Jan; 110(15):60–8. 6. Nghe NT, Janecek P, Haddawy P. A comparative analysis of techniques for predicting academic performance. 37th An- 5. Conclusion nual Frontiers in Education Conference-Global Engineer- ing: Knowledge without Borders, Opportunities without This paper discussed the work done in educational data Passports 2007 Oct, T2G-7, 2007. mining categorically in traditional education. Within each 7. Ogor EN. Student academic performance monitoring and category the works are again discussed chronologically. evaluation using data mining techniques. Electronics, Ro- botics and Automotive Mechanics Conference, 2007 Sep. p. We consider research areas of traditional education, as 354–9. Crossref our study involves traditional education. 8. Oladokun VO, Adebanjo AT, Charles-Owaba OE. Predict- In traditional education, performance prediction ing students’ academic performance using artificial neural is in matured state with contribution from many network: A case study of an engineering course. The Pacific researchers. However, there is paucity of research in the Journal of Science and Technology. 2008 May; 9(1):72–9. 9. Cortez P, Silva AM. Using data mining to predict secondary field of employability prediction. As both performance school student performance. 2008; 1–8. and employability of students graduating from an 10. Ramaswami M, Bhaskaran R. A study on feature selection institution decide the market value of the institution, techniques in educational data mining. ArXiv preprint arX- research is required to develop comprehensive models iv. 2009 Dec; 1(1):7–11. for performance and employability tool and develop a 11. Paris IH, Aen ff dey LS, Mustapha N. Improving academ- ic performance prediction using voting technique in data system that will be able to predict both performance and mining. World Academy of Science, Engineering and Tech- employability. From the literature review, it is clear that nology. 2010 Feb; 62:820–3. most commonly used predictors are socio economic / 12. Kovacic Z. Early prediction of student success: Mining stu- demographic profile and past academic record of the dents’ enrolment data. 2010; 647–65. students. Apart from this, number of hours dedicated 13. Kovačić ZJ, Green JS. Predictive working tool for early to studies, distance of the institution from home, loan, identification of ‘at risk’students. 2010; 1–73. 14. Ramaswami M, Bhaskaran R. A CHAID based perfor- internet facility etc. has been considered by the individual mance prediction model in educational data mining. ArXiv researchers in their studies. Thus in general researchers preprint arXiv. 2010 Feb; 7(1):1–9. in the field of EDM have focused on academic and 15. Guruler H, Istanbullu A, Karahasan M. A new student social integration of students for performance and performance analysing system using knowledge discovery employability prediction. The effect of emotional skills in higher educational databases. Computers & Education. 2010 Aug; 55(1):247–54. Crossref on academic performance and employability needs to 16. Delen D. A comparative analysis of machine learning tech- be explored further. The future work includes survey of niques for student retention management. Decision Sup- tools, available for prediction of academic performance port Systems. 2010 Nov; 49(4):498–506. Crossref and employability. 17. Aen ff dey LS, Paris IH, Mustapha N, Suleiman MN, Muda Z. Ranking of influencing factors in predicting students’ academic performance. Information Technology Journal. 6. References 2010; 9(4):832–7. Crossref 18. Baradwaj BK, Pal S. Mining educational data to analyze stu- 1. Romero C, Ventura S. Educational data mining: a review of dents’ performance. ArXiv preprint arXiv: 1201.3417. 2012 the state of the art. IEEE Transactions on Systems, Man, and Jan; 2(6):63–9. Cybernetics, Part C (Applications and Reviews). 2010 Nov; 19. Huang S. Predictive modeling and analysis of student ac- 40(6):601–18. Crossref ademic performance in an engineering dynamics course. 2. Baker RS, Yacef K. The state of educational data mining in 2011; 1–136. Vol 10 (24) | June 2017 | www.indjst.org Indian Journal of Science and Technology Students’ Performance and Employability Prediction through Data Mining: A Survey 20. Cheewaprakobkit P. Study of Factors Analysis Ae ff cting Ac- challenges via curriculum design and restructuring. Inter- ademic Achievement of Undergraduate Students in Inter- national Conference on Electrical Engineering and Infor- national Program. Proceedings of the International Multi matics. 2009 Aug; 2:357–61. Crossref Conference of Engineers and Computer Scientists. 2013; 32. Shafie LA, Nayan S. Employability awareness among Ma- 1:13–5. laysian undergraduates. International Journal of Business 21. Şen B, Uçar E, Delen D. Predicting and analyzing secondary and Management. 2010 Aug; 5(8):119–23. education placement-test scores: A data mining approach. 33. Gokuladas VK. Predictors of employability of engineering Expert Systems with Applications. 2012 Aug; 39(10):9468– graduates in campus recruitment drives of Indian software 76. Crossref services companies. International Journal of Selection and 22. Osmanbegović E, Suljić M. Data mining approach for pre- Assessment. 2011 Sep; 19(3):313–9. Crossref dicting student performance. Economic Review. 2012 May; 34. Othman Z, Musa F, Mokhtar NH, Ya’acob A, Latiff RA, 10(1):3–12 Hussin H. Investigating University Graduates’ English Lan- 23. Shah NS. Predicting Factors That Ae ff ct Students’ academic guage Competency towards Employability: A Proposed Performance by Using Data Mining Techniques. Pakistan Research Method. International Journal of Learning. 2010 business review. 2012 Jan; 631(68):631–810. Aug; 17(7):429–40. 24. Kabakchieva D. Predicting student performance by using 35. Potgieter I, Coetzee M. Employability attributes and per- data mining methods for classification. Cybernetics and in- sonality preferences of postgraduate business management formation technologies. 2013 Mar; 13(1):61–72. Crossref students. SA Journal of Industrial Psychology. 2013 Jan; 25. Ramesh VA, Parkavi P, Ramar K. Predicting student perfor- 39(1):1–10. Crossref mance: a statistical and data mining approach. Internation- 36. Yusoff YM, Omar MZ, Zaharim A, Mohamed A, Muhamad al journal of computer applications. 2013 Jan; 63(8):35–9. N. Employability skills performance score for fresh engi- Crossref neering graduates in Malaysian industry. Asian Social Sci- 26. Harvey L. Defining and measuring employability. Quality ence. 2012 Dec; 8(16):140–5. Crossref in higher education. 2001 Jul; 7(2):97–109. Crossref 37. Jantawan B, Tsai CF. The Application of Data Mining to 27. Gokuladas VK. Technical and nontechnical education and Build Classification Model for Predicting Graduate Em- the employability of engineering graduates: an Indian case ployment. ArXiv preprint arXiv: 1312.7123. 2013 Dec; study. International Journal of Training and Development. 11(10):1–7. 2010 Jun; 14(2):130–43. Crossref 38. Pool LD, Qualter P. Emotional self‐efficacy, graduate em- 28. Rees C, Forbes P, Kubler B. Student employability profiles. ployability, and career satisfaction: Testing the associations. New York: The Higher Education Academy; 2006. p. 1–83. Australian Journal of Psychology. 2013 Dec; 65(4):214–23. 29. Kahya E. The effects of job performance on effectiveness. Crossref International Journal of Industrial Ergonomics. 2009 Jan; 39. Mishra T, Kumar D, Gupta S. Mining Students’ Data for 39(1):96–104. Crossref Prediction Performance. Fourth International Conference 30. Chien CF, Chen LF. Data mining to improve personnel on Advanced Computing & Communication Technologies. selection and enhance human capital: A case study in 2014 Feb; 255–62. Crossref high-technology industry. Expert Systems with applica- 40. Mishra T, Kumar D, Gupta S. Students’ Employability Pre- tions. 2008 Jan; 34(1):280–90. Crossref diction Model through Data Mining. International Journal 31. Mukhtar M, Yahya Y, Abdullah S, Hamdan AR, Jailani N, of Applied Engineering Research. 2016; 11(4):2275–82. Abdullah Z. Employability and service science: Facing the Vol 10 (24) | June 2017 | www.indjst.org Indian Journal of Science and Technology http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Indian Journal of Science and Technology Unpaywall

Loading next page...
 
/lp/unpaywall/students-performance-and-employability-prediction-through-data-mining-Vuhx7JIvOo

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
Unpaywall
ISSN
0974-5645
DOI
10.17485/ijst/2017/v10i24/110791
Publisher site
See Article on Publisher Site

Abstract

Objective: To systematically review the work done in the field of academic performance prediction and employability prediction of students in higher education. Methods: The survey first explain show higher education has become an exciting field of research and why the prediction of academic performance and employability is beneficial for the institutions. We also explain briefly in how many ways higher education is being provided world-wide. Then we discuss the work done in both the areas of prediction. Findings: The survey explores existing research highlights and finds that prediction of academic performance has progressed a lot but employability prediction is yet to mature. Application: It further suggests few parameters that have not been considered so far in predicting the performance or employability. Keywords: Academic Performance, Data Mining, Employability, Higher Education, Prediction, Survey e r Th eview is organized chronologically and 1. Introduction categorically to oer in ff sight on how past research efforts laid the groundwork for subsequent studies, including In the present knowledge-based epoch, education plays the present research efforts. The detailed review has been a major role in the progress of a nation’s economy and carried out, so that; the present research can properly be development. It assures to invigorate the country by tailored to add the present body of literature, as well as the contributing the reliable and quality workforce to the scope and direction of the present research effort. society. Higher education is the foundation for fostering In order to give a comprehensive view of the work the talent, the key factor in increasing national human done in the field of EDM, many survey papers have been capital quality, and the main way to upgrade a nation’s published. The most cited research compiled the work competitive status. Thus, the research on development done till 2005 and the other that has discussed vital of higher education is an important work and is actually features of EDM . The work done in EDM till 2010, 2013, required. To get an edge over each other, institutions are 3–5 2014 has been complied in the different research papers applying cutting edge technologies like data mining on (Table 1). the huge data generated in class room including academic, According to the literature study, EDM research behavioral, demographic data of students and faculty pertains mainly three heads, according to the way data data as well. e Th data generated in educational set up can is collected- give deep insight into educational process. Educational • Traditional face to face or the offline education Data Mining (EDM) refers mining the data generated in system based on data generated in the classroom. educational set up. • E-learning in which the learning is provided through Researches in EDM have benefited the educational online content based on online activity logs. setups tremendously. * Author for correspondence Students’ Performance and Employability Prediction through Data Mining: A Survey • Intelligent tutoring system (ITS) and Adaptive 3. Prediction of Students’ Educational Hypermedia System (AEHS) involve Performance online teaching based on students need, his or her progress rather than providing same structured Students’ academic performance is a mature field now lesson to all the students. with many researchers contributing to it. Moving in chronological increasing order from 2007 onwards the Table 1. Survey Papers in Educational Data Mining researchers have considered various parameters for References Highlights prediction and using different algorithm they have tried Comprehensive survey for traditional to predict the result of specific course of a particular educational systems. university. Highlights the work done in EDM from the Authors have considered students from two different view point of models, emergence of public data, tools. institutions, one International and one small institution, Illustrates problems of education system to check the effect of research separately on both. In both resolved using data mining and proposed the the cases decision tree provided better accuracy than association of techniques for better predictions. Bayesian Network. Recognized different educational system tasks, A simple student performance assessment and discipline, techniques, and algorithms. monitoring system based on various data mining Provides a comprehensive survey, a travelogue techniques was developed with the predictor attributes from 2002 to 2014 for educational data mining. including students’ demographic details, course average st th 7 score in 1 to 5 semester overall gain performance, etc . E learning and Intelligent Tutoring systems have used Decision Tree C5 showed highest accuracy, followed by EDM mainly to model online behavior of students, track Classification and Regression, Trees (CART), Artificial their performance and get feedback from them. Neural Network (ANN), Chi-Squared Automatic However, this survey explores traditional educational Interaction Detection (CHAID). settings as it is still most widely used method of teaching Taking more factors like university matriculation across the globe. exam, GCE (General Certificate of Education) Score, Senior Secondary Certified Examination score SSCE), grades in O level subject, location of University from 2. Traditional Educational Setup home, gender and age, Cumulative Grade point was to Traditional educational set up refers to class room teaching classified as Good, Average and Poor. Artificial Neural which is still the most popular form of teaching by the Network (ANN) was used for prediction and 74.5 Institutions across the globe and needs to be explored. %accuracy was attained in performance prediction . Students prefer to take admission in an Institution with Prediction of school students’ performance has high academic performance and where the passing out been considered with a total of 33 parameters including students have better employability. Prediction of both socio -demographic details like (parental marital status, academic performance and employability can help the father’s job, mother’s job, quality of family relationship, management identify students at risk of poor academic attitude towards study (No. of hour, past failure) Internet performance and low employability. The process of facility, family support, free time aer s ft chool, health, prediction involves application of various data mining alcohol consumption etc . To predict the performance, algorithms, to predict the dependent variables based on Decision was used. Attendance, parents job, previous year independent factors. performance was found to be the key factors that ae ff ct Next two subsections discuss the researcher’s the current achievement. inclination towards prediction of both students’ e Th study uses feature selection process of Waikato performance and their employability. The focus is on Environment for Knowledge Analysis (WEKA) tool the attributes considered and methods/classification which has inbuilt set of methods and considers Student’s algorithms adopted for prediction of performance and gender, Eyesight, Community, Physical handicap, food employability prediction. habit, family details, mode of transport, medium of Vol 10 (24) | June 2017 | www.indjst.org Indian Journal of Science and Technology Tripti Mishra, Dharminder Kumar and Sangeeta Gupta instruction, sports activity etc. as predictive factors.It consider demographical details of the students, but was also observed that classification methods like Naïve concentrate only on grades of various courses as attributes Bayes, one R voted perception performed much better to predict categories of students as first class, the other with feature selected subset than where all variables were consisting second class upper lower and third class. In this considered. case Naive Bayes was found to show better accuracy than Voting technique is a method of applying more than Decision Tree. In a contrast authors have used previous one algorithm in succession and then taking the best academic achievements in exams and present academic result. e Th researchers have concluded that Decision assessment in Lab and class assignments to predict end Stump along with Hidden Naïve Bayes is most suited for semester marks, whereas another research paper has academic performance prediction in case of a New Zealand made academic performance prediction of Engineering Polytechnic, where, apart from academic performance, Drawing course. SVM is well suited for individual demography, disability was also considered as predictive students’ performance prediction and regression is to be factors. In a similar research CART algorithm has used for prediction of entire class. 13 20 been found to give highest accuracy. Further, the study In another study language proficiency in English, added few more predictive parameters like ethnicity, and credit selection and whether a student is unmarried, student’s current job condition to predict performance. In married, divorced have shown their effect on academic this case tree was found less accurate than regression and performance prediction and decision tree showed better analysis. accuracy over neural network. Chi Squared Automatic Interaction Detector eTh research used various factors that were (CHAID) was used make high school result prediction ranked according to their effectiveness in predicting with predicting factors being taken as health of the the placement result of Turkish secondary school. The student, tuition availability, facilities to study at home etc. predictors included marks Spirituality and morals, Prediction accuracy was not very good 44.69% and key Turkish language and entry level exam. Decision tree influences were found to be mother’s education, location worked well with the problem. from school etc . Demographic variable, score of high school entrance The study aimed to discover individual characteristics exam and attributes related to their attitude towards that decide their success using Microsoft Decision studying etc. were considered for predicting the result Tree.11 attributes that includes registered information of 1st year students of economics course . A set of four high school information; Turkish, University entrance algorithms were applied. It was found that past academic exams degree and University placement info family living records, entrance exam marks, hours put into studies are conditions and financial status etc. were considered. having maximum impact on prediction accuracy. Microsoft Decision Tree (MDT) has been used to predict Students’ academic performance in five categories GPA with just two categories successful and unsuccessful. were predicted using Decisions Tree, Functions, Rules, Four classification methods Artificial Neural Network, Bayes Net etc. and Random tree provided the best Decision trees, Support Vector Machine and logistic result. In a more recent study , factors considered were regression along with ensemble techniques (i.e. bagging, demographic profile, previous academic scores, entrance busting and information fusion) to analyze 16,000 exam result and among the various algorithms applied j48 students records . 39 variables including demographic gave highest accuracy of prediction. data, TOEFL, SAT Score, Loan etc. were used to predict In yet another research paper at school level attrition/ retention. Support Vector machine provided performance was predicted using Parental status, Mode Highest Accuracy followed by decision tree, Artificial of transport, Groups of subjects, type of school, previous Neural network and regression. marks and applying algorithms decision tree, KNN, SVM In another research paper attribute importance was etc. emphasized by ranking them, using correlation based Academic performance is always one of the primary feature subset selection and consistency subset selection predictors of employability. Thus all the factors ae ff cting (COE) and using them further find accuracy of various academic performance also ae ff ct the employability. Next classifiers . Unlike other researchers, authors do not we explore the work done on employability. Vol 10 (24) | June 2017 | www.indjst.org Indian Journal of Science and Technology Students’ Performance and Employability Prediction through Data Mining: A Survey were stronger predictor of their employability than 4. Prediction of Students’ their technical education consisting of their academic Employability performance . Poor English language competency is found to be a major reason for the low employability . Today, the reputation of an Institution is judged by This has been supported by one more study which its academic success, its ability to retain students concludes that knowledge of GPA and English Language and to provide employment for its students. The competency are required for the students in software term “Employability” still has no precise definition. industry to continue with their employment and also Employability has been described in many ways, like, the the female candidates were better performer than male ability to secure a job, getting a job within a specified time candidates in campus placement drives. period aer g ft raduating, the ability to skill map oneself Psychology seeks an answer to the kind of skill according to the job need, or the willingness of the student requirement which is essential for enhancing job to extend the graduate learning at work . Alternatively, prospects. A sound effect of overall personality of the employability is defined as the ability of students to employee along with career competence, self-efficacy is secure a job during on campus placements . Research seen on employability. This has been supported by the in employability prediction is in nascent stage. It mostly study where employers give priority to soft skills, team involves identification of skills or attributes required work, etc. from the perspective of employers and is obtained from Another researcher concludes that J48 is most employer through questionnaire and interviews. Mostly suited algorithm to predict the employability based statistical methods have been applied and research is on demographic profile, employees job satisfaction, more of descriptive than predictive. academics etc. e im Th portance of psychological factors along It is comprehensible that not much work has been with personal and organizational awareness has been done in the direction of employability prediction emphasized in a report by Higher Education Academy mainly due to lack of authentic data, hence this study of with the Council for Industry and Higher Education employability prediction and model development will (CIHE) in United Kingdom another study considered contribute significantly in educational data mining. the effect of working environment on performance of the Graduate employability has been the subject of little employee . empirical research. er Th e are a number of difficulties Whether an employee will be able to meet the in defining and measuring graduate employability, expectation and should be hired is predicted in research which means that there is a paucity of research that paper which is based on few attributes extracted from looks at its predictors and outcomes. Previous work has candidate’s curriculum vital key words in application and proposed that emotional competence improves graduate interview. employability. Researchers in psychology have shown It has been concluded that as the paradigm is shifting that the emotional skills of a student are also important from product based to service based industry especially factors for performance prediction. However not much in Information Technology, and hence the curriculum work has been done in the field of EDM to validate the and method of delivering lecture must evolve in order to existing knowledge or construct new knowledge about enhance employability . emotional skill being predictor of performance. One A descriptive study in the research indicates that reason for not considering emotional skill is lack of employers look forward to employees with Personal authentic data. The need of hour is to construct authentic attributes that include loyalty, commitment, honesty, primary data that has factors of academic integration integrity, enthusiasm, reliability, personal presentation, and social integration and emotional skills and study the common sense, positive self-esteem, and a sense of humor, prediction of academic performance and employability motivation, adaptability, a balanced attitude towards work in tandem. e Th authors of this paper have identified this and home life and ability to deal with pressure. gap and published two papers in this regard where apart Correlation and ordinal regression has been used from social and academic integration, emotional skills to conclude that a students’ non-technical education like leadership, self-esteem, empathy; decision making consisting of reasoning, logical ability, and soft skills Vol 10 (24) | June 2017 | www.indjst.org Indian Journal of Science and Technology Tripti Mishra, Dharminder Kumar and Sangeeta Gupta 2009: A review and future visions. JEDM-Journal of Educa- capability, time management and stress management tional Data Mining. 2009 Oct; 1(1):3–17 are also considered as predictor variables. A model 3. Romero C, Ventura S. Educational data mining: A survey was derived for performance prediction based on from 1995 to 2005. Expert systems with applications. 2007 academic, social and above emotional skills. Similarly, Jul; 33(1):135–46. Crossref Employability prediction model has been developed by 4. Pe-a-Ayala A. Educational data mining: A survey and a data the authors using authentic primary data that involves mining-based analysis of recent works. Expert systems with applications. 2014 Mar; 41(4):1432–62. Crossref social, academic as well as emotional skills. Future work 5. a Th kar P. Performance Analysis and Prediction in Educa- requires developing tools for prediction of performance tional Data Mining: A Research Travelogue. International and employability. Journal of Computer Applications. 2015 Jan; 110(15):60–8. 6. Nghe NT, Janecek P, Haddawy P. A comparative analysis of techniques for predicting academic performance. 37th An- 5. Conclusion nual Frontiers in Education Conference-Global Engineer- ing: Knowledge without Borders, Opportunities without This paper discussed the work done in educational data Passports 2007 Oct, T2G-7, 2007. mining categorically in traditional education. Within each 7. Ogor EN. Student academic performance monitoring and category the works are again discussed chronologically. evaluation using data mining techniques. Electronics, Ro- botics and Automotive Mechanics Conference, 2007 Sep. p. We consider research areas of traditional education, as 354–9. Crossref our study involves traditional education. 8. Oladokun VO, Adebanjo AT, Charles-Owaba OE. Predict- In traditional education, performance prediction ing students’ academic performance using artificial neural is in matured state with contribution from many network: A case study of an engineering course. The Pacific researchers. However, there is paucity of research in the Journal of Science and Technology. 2008 May; 9(1):72–9. 9. Cortez P, Silva AM. Using data mining to predict secondary field of employability prediction. As both performance school student performance. 2008; 1–8. and employability of students graduating from an 10. Ramaswami M, Bhaskaran R. A study on feature selection institution decide the market value of the institution, techniques in educational data mining. ArXiv preprint arX- research is required to develop comprehensive models iv. 2009 Dec; 1(1):7–11. for performance and employability tool and develop a 11. Paris IH, Aen ff dey LS, Mustapha N. Improving academ- ic performance prediction using voting technique in data system that will be able to predict both performance and mining. World Academy of Science, Engineering and Tech- employability. From the literature review, it is clear that nology. 2010 Feb; 62:820–3. most commonly used predictors are socio economic / 12. Kovacic Z. Early prediction of student success: Mining stu- demographic profile and past academic record of the dents’ enrolment data. 2010; 647–65. students. Apart from this, number of hours dedicated 13. Kovačić ZJ, Green JS. Predictive working tool for early to studies, distance of the institution from home, loan, identification of ‘at risk’students. 2010; 1–73. 14. Ramaswami M, Bhaskaran R. A CHAID based perfor- internet facility etc. has been considered by the individual mance prediction model in educational data mining. ArXiv researchers in their studies. Thus in general researchers preprint arXiv. 2010 Feb; 7(1):1–9. in the field of EDM have focused on academic and 15. Guruler H, Istanbullu A, Karahasan M. A new student social integration of students for performance and performance analysing system using knowledge discovery employability prediction. The effect of emotional skills in higher educational databases. Computers & Education. 2010 Aug; 55(1):247–54. Crossref on academic performance and employability needs to 16. Delen D. A comparative analysis of machine learning tech- be explored further. The future work includes survey of niques for student retention management. Decision Sup- tools, available for prediction of academic performance port Systems. 2010 Nov; 49(4):498–506. Crossref and employability. 17. Aen ff dey LS, Paris IH, Mustapha N, Suleiman MN, Muda Z. Ranking of influencing factors in predicting students’ academic performance. Information Technology Journal. 6. References 2010; 9(4):832–7. Crossref 18. Baradwaj BK, Pal S. Mining educational data to analyze stu- 1. Romero C, Ventura S. Educational data mining: a review of dents’ performance. ArXiv preprint arXiv: 1201.3417. 2012 the state of the art. IEEE Transactions on Systems, Man, and Jan; 2(6):63–9. Cybernetics, Part C (Applications and Reviews). 2010 Nov; 19. Huang S. Predictive modeling and analysis of student ac- 40(6):601–18. Crossref ademic performance in an engineering dynamics course. 2. Baker RS, Yacef K. The state of educational data mining in 2011; 1–136. Vol 10 (24) | June 2017 | www.indjst.org Indian Journal of Science and Technology Students’ Performance and Employability Prediction through Data Mining: A Survey 20. Cheewaprakobkit P. Study of Factors Analysis Ae ff cting Ac- challenges via curriculum design and restructuring. Inter- ademic Achievement of Undergraduate Students in Inter- national Conference on Electrical Engineering and Infor- national Program. Proceedings of the International Multi matics. 2009 Aug; 2:357–61. Crossref Conference of Engineers and Computer Scientists. 2013; 32. Shafie LA, Nayan S. Employability awareness among Ma- 1:13–5. laysian undergraduates. International Journal of Business 21. Şen B, Uçar E, Delen D. Predicting and analyzing secondary and Management. 2010 Aug; 5(8):119–23. education placement-test scores: A data mining approach. 33. Gokuladas VK. Predictors of employability of engineering Expert Systems with Applications. 2012 Aug; 39(10):9468– graduates in campus recruitment drives of Indian software 76. Crossref services companies. International Journal of Selection and 22. Osmanbegović E, Suljić M. Data mining approach for pre- Assessment. 2011 Sep; 19(3):313–9. Crossref dicting student performance. Economic Review. 2012 May; 34. Othman Z, Musa F, Mokhtar NH, Ya’acob A, Latiff RA, 10(1):3–12 Hussin H. Investigating University Graduates’ English Lan- 23. Shah NS. Predicting Factors That Ae ff ct Students’ academic guage Competency towards Employability: A Proposed Performance by Using Data Mining Techniques. Pakistan Research Method. International Journal of Learning. 2010 business review. 2012 Jan; 631(68):631–810. Aug; 17(7):429–40. 24. Kabakchieva D. Predicting student performance by using 35. Potgieter I, Coetzee M. Employability attributes and per- data mining methods for classification. Cybernetics and in- sonality preferences of postgraduate business management formation technologies. 2013 Mar; 13(1):61–72. Crossref students. SA Journal of Industrial Psychology. 2013 Jan; 25. Ramesh VA, Parkavi P, Ramar K. Predicting student perfor- 39(1):1–10. Crossref mance: a statistical and data mining approach. Internation- 36. Yusoff YM, Omar MZ, Zaharim A, Mohamed A, Muhamad al journal of computer applications. 2013 Jan; 63(8):35–9. N. Employability skills performance score for fresh engi- Crossref neering graduates in Malaysian industry. Asian Social Sci- 26. Harvey L. Defining and measuring employability. Quality ence. 2012 Dec; 8(16):140–5. Crossref in higher education. 2001 Jul; 7(2):97–109. Crossref 37. Jantawan B, Tsai CF. The Application of Data Mining to 27. Gokuladas VK. Technical and nontechnical education and Build Classification Model for Predicting Graduate Em- the employability of engineering graduates: an Indian case ployment. ArXiv preprint arXiv: 1312.7123. 2013 Dec; study. International Journal of Training and Development. 11(10):1–7. 2010 Jun; 14(2):130–43. Crossref 38. Pool LD, Qualter P. Emotional self‐efficacy, graduate em- 28. Rees C, Forbes P, Kubler B. Student employability profiles. ployability, and career satisfaction: Testing the associations. New York: The Higher Education Academy; 2006. p. 1–83. Australian Journal of Psychology. 2013 Dec; 65(4):214–23. 29. Kahya E. The effects of job performance on effectiveness. Crossref International Journal of Industrial Ergonomics. 2009 Jan; 39. Mishra T, Kumar D, Gupta S. Mining Students’ Data for 39(1):96–104. Crossref Prediction Performance. Fourth International Conference 30. Chien CF, Chen LF. Data mining to improve personnel on Advanced Computing & Communication Technologies. selection and enhance human capital: A case study in 2014 Feb; 255–62. Crossref high-technology industry. Expert Systems with applica- 40. Mishra T, Kumar D, Gupta S. Students’ Employability Pre- tions. 2008 Jan; 34(1):280–90. Crossref diction Model through Data Mining. International Journal 31. Mukhtar M, Yahya Y, Abdullah S, Hamdan AR, Jailani N, of Applied Engineering Research. 2016; 11(4):2275–82. Abdullah Z. Employability and service science: Facing the Vol 10 (24) | June 2017 | www.indjst.org Indian Journal of Science and Technology

Journal

Indian Journal of Science and TechnologyUnpaywall

Published: Jun 1, 2017

There are no references for this article.