Predicting Academic Performance Based on Optical Mark Recognition Patterns: A Comparative Study of Transfer Learning Models

Öz

Optical forms, which are widely used assessment tools in education, have the potential to reflect students' cognitive, sensory, and behavioral characteristics. This study aims to go beyond the traditional use of Optical Mark Recognition (OMR) systems by examining students’ marking behaviors as a predictor of academic achievement. Using 2,100 marking images collected from 42 participants (21 per class), we evaluated 18 transfer learning–based feature extractors, of which 17 were successfully implemented, in combination with 25 classification algorithms. To eliminate potential data leakage and ensure generalization to unseen individuals, all experiments were conducted using 10-fold GroupKFold cross-validation with subject-wise splitting, such that all samples from the same participant were kept within the same fold. The best-performing configuration, EfficientNet-B0 feature representations combined with Support Vector Classification, achieved 88.90% accuracy, with strong threshold-independent performance (ROC-AUC = 0.9268; PR-AUC = 0.8927; Average Precision = 0.8934). Statistical validation via the Friedman test (χ²(16) = 174.34, p < .001) confirmed significant performance differences across transfer learning architectures. These findings indicate that markings on optical forms should not be treated as random artifacts but as behavioral traces that reflect underlying cognitive and affective processes, and they support a shift from a results-oriented to a process-oriented assessment paradigm. From a learning analytics and educational policy perspective, the proposed approach positions paper-based OMR sheets as low-cost “behavior sensors” that can complement early warning mechanisms by enabling earlier identification of students at academic risk and facilitating timely, targeted interventions.

Anahtar Kelimeler

• Academic Achievement Prediction
• Early Warning Systems
• Optical Signal Recognition

Optik İşaret Tanıma Kalıplarına Dayalı Akademik Performans Tahmini: Transfer Öğrenme Modellerinin Karşılaştırmalı Bir Çalışması

Öz

Eğitimde yaygın olarak kullanılan değerlendirme araçları olan optik formlar, öğrencilerin bilişsel, duyusal ve davranışsal özelliklerini yansıtma potansiyeline sahiptir. Bu çalışma, öğrencilerin işaretleme davranışlarını akademik başarının bir öngörücüsü olarak inceleyerek, Optik İşaret Tanıma (OMR) sistemlerinin geleneksel kullanımının ötesine geçmeyi amaçlamaktadır. 42 katılımcıdan (sınıf başına 21) toplanan 2.100 işaretleme görüntüsünü kullanarak, 18 transfer öğrenme tabanlı özellik çıkarıcıyı değerlendirdik ve bunların 17'si 25 sınıflandırma algoritmasıyla birlikte başarıyla uygulandı. Olası veri sızıntısını ortadan kaldırmak ve görülmemiş bireylere genelleme sağlamak için, tüm deneyler, aynı katılımcıdan alınan tüm örneklerin aynı kat içinde tutulacağı şekilde, konu bazında bölünmüş 10 katlı GroupKFold çapraz doğrulama kullanılarak gerçekleştirilmiştir. En iyi performans gösteren yapılandırma, Support Vector Classification ile birleştirilen EfficientNet-B0 özellik temsilleri, %88,90 doğruluk oranına ulaştı ve eşikten bağımsız güçlü bir performans gösterdi (ROC-AUC = 0,9268; PR-AUC = 0,8934). Friedman testi (χ²(16) = 174,34, p < .001) ile yapılan istatistiksel doğrulama, transfer öğrenme mimarileri arasında önemli performans farklılıkları olduğunu doğruladı. Bu bulgular, optik formlardaki işaretlerin rastgele artefaktlar olarak değil, altta yatan bilişsel ve duyusal süreçleri yansıtan davranış izleri olarak değerlendirilmesi gerektiğini göstermektedir ve sonuç odaklı bir değerlendirme paradigmasından süreç odaklı bir değerlendirme paradigmasına geçişi desteklemektedir. Öğrenme analitiği ve eğitim politikası perspektifinden bakıldığında, önerilen yaklaşım, kağıt tabanlı OMR sayfaları, akademik risk altındaki öğrencilerin daha erken tespit edilmesini sağlayarak ve zamanında, hedefe yönelik müdahaleleri kolaylaştırarak erken uyarı mekanizmalarını tamamlayabilen düşük maliyetli “davranış sensörleri” olarak konumlandırmaktadır.

Anahtar Kelimeler

• Akademik Başarı Tahmini
• Erken Uyarı Sistemleri
• Optik İşaret Tanıma

Kaynakça

1. Luciano RG. Innovative test item analysis using optical mark recognition technology: An evaluation. Int J Adv Appl Sci 2025; 12: 1–11.
2. Sievertsen HH. Assessments in education. In: Oxford Research Encyclopedia of Economics and Finance. Oxford: Oxford University Press, 2023.
3. Marshall P. Contribution of open-ended questions in student evaluation of teaching. High Educ Res Dev 2022; 41: 1992–2005.
4. Walstad WB, Saunders P. The Principles of Economics Course: A Handbook for Instructors. New York, NY, USA: McGraw-Hill College, 1990.
5. Contreras JO, Hilles S, Bakar ZA. Essay question generator based on Bloom’s taxonomy for assessing automated essay scoring system. In: Proc 2nd Int Conf Smart Computing and Electronic Enterprise; 2021; IEEE. pp. 55–62.
6. Birgili B. Open ended questions as an alternative to multiple choice: Dilemma in Turkish examination system. 2014.
7. Azevedo J, Oliveira EP, Beites PD. E-assessment and multiple-choice questions. In: Advances in Educational Technologies. 2019. pp. 1–27.
8. Ali K, Zahra D. Ten tips for effective use and quality assurance of multiple-choice questions in knowledge-based assessments. Eur J Dent Educ 2024; 28: 655–662.

9. Das B, Majumder M, Phadikar S, Sekh AA. Multiple-choice question generation with auto-generated distractors for computer-assisted educational assessment. Multimed Tools Appl 2021; 80: 31907–31925.
10. Karunanayake N. OMR sheet evaluation by web camera using template matching approach. 2023.
11. Hasan RH, Aboud IS, Hassoon RM, Khioon ASA. Optical mark recognition using modify bi-directional associative memory. Tikrit J Pure Sci 2024; 29: 174–184.
12. Singh JK, Kulkarni S, Patil SB, et al. OMR automated grading. Int J Innov Sci Res Technol 2024; 3757–3761.
13. Jain V, Malik S, Bhatia V. Robust image processing based real-time optical mark recognition system. In: Proc IEEE 6th Conf Inf Commun Technol; 2022.
14. Kakade N, Jaiswal RC. OMR sheet evaluation using image processing. 2017.
15. Afifi M, Hussain KF. Flexibility in multiple-choice-based tests using image classification techniques. Int J Doc Anal Recognit 2019; 22: 127–142.
16. Vetrivel S, Vidhyapriya P, Arun VP. The role of AI in transforming assessment practices in education. 2024. pp. 43–70. [17] Zimmerman BJ. Becoming a self-regulated learner: An overview. Columbus, OH, USA: Ohio State University Press, 2002.
17. Sopariwala S, Kasat D. Handwriting analysis and personality profiling using image processing and machine learning. In: Multifaceted Approaches for Data Acquisition Processing and Communication. Boca Raton, FL, USA: CRC Press, 2024. pp. 111–117.
18. Alalawi K, Chiong R, Athauda R. Early detection of under-performing students using machine learning algorithms. In: Proc IEEE CITISIA; 2021.
19. Stark JT. Enhancing algorithmic early warning systems with dynamic selection to predict high school graduation outcomes. 2024.
20. Aguilar S, Lonn S, Teasley SD. Perceptions and use of an early warning system during a higher education transition program. In: Proc 4th Int Conf Learning Analytics and Knowledge (LAK ’14); 2014. pp. 113–117. doi:10.1145/2567574.2567625. In: Proc ACM Conf; 2014.
21. Agduk S, Aydemir E. Classification of handwritten text signatures by person and gender: A comparative study of transfer learning methods. 2022.
22. Chansky NM. A note on the grade point average in research. Educ Psychol Meas 1964; 24: 95–99.
23. Lucio R, Hunt E, Bornovalova M. Predicting academic failure with risk factors. Dev Psychol 2012; 48: 422–428.
24. Dwivedi DN, Mahanty G, Dwivedi VN. Predictive analytics in personalized education. In: Enhancing Education With Intelligent Systems. IGI Global, 2024. pp. 44–59.
25. Frederick S. Cognitive reflection and decision making. J Econ Perspect 2005.
26. Zeidner M. Test Anxiety: The State of the Art. Dordrecht: Kluwer Academic, 1998.
27. Wang D. Educational data mining: Methods and applications. Appl Comput Eng 2023; 16: 205–209.
28. Khodeir N. Student modeling using educational data mining techniques. In: Proc ACCS/PEIT; 2019. pp. 7–14.
29. Santos GAS, Bordignon AL, Oliveira SLG, et al. Educational data mining for dropout prediction. 2019. pp. 86–91.
30. Bellaj M, Ben Dahmane A, Sefian L. Machine learning for student performance prediction. Int J Online Biomed Eng 2024; 20: 55–74.
31. Mutrofin S, Maisarah M, Widodo S, et al. PCA and logistic regression based SVM for EDM. In: IOP Conf Ser Mater Sci Eng; 2020.
32. Naicker N, Adeliyi T, Wing J. Linear SVM for student performance prediction. Math Probl Eng 2020.
33. Dangi A, Srivastava S. Educational data classification using selective Naïve Bayes. In: Proc IEEE MITE; 2014. pp. 118–121.
34. Chen Y, Jin K. Educational performance prediction with random forest. Int J Adv Comput Sci Appl 2024; 15: 69–78.
35. Luo Y, Wang Z. Feature mining via interpretable deep neural networks. In: Proc ICIET; 2024.
36. Hernández-Blanco A, Herrera-Flores B, Tomás D, Navarro-Colorado B. Deep learning in EDM: A systematic review. 2019.
37. Khan M, Naz S, Khan Y, et al. Student performance prediction from LMS logs. IEEE Access 2023; 11: 86953–86962.
38. Ajibade SSM, Ahmad NBB, Shamsuddin SM. Ensemble methods for student performance modeling. In: IOP Conf Ser Mater Sci Eng; 2019.
39. Alyahyan E, Düştegör D. Predicting academic success in higher education. 2020.
40. Malik S, Malik S. Hybrid data science approaches for academic performance prediction. In: Lect Notes Electr Eng. Springer, 2024. pp. 521–539.
41. Mengarelli L, Kostiuk B, Vitório JG, et al. OMR metrics and evaluation: A systematic review. Multimed Tools Appl 2020; 79: 6383–6408.
42. Jingyi T, Hooi YK, Bin OK. Image processing for enhanced OMR precision. In: Proc ICCOINS; 2021. pp. 322–327.
43. Tümer AE, Küçükkara Z. Image processing–oriented OMR evaluation system. Int J Appl Math Electron Comput 2018; 6: 59–64.
44. Rosca CM. Comparative analysis of object classification algorithms. Rom J Pet Gas Technol 2023; 75: 169–180.
45. R. C. . Dharmik, S. . Rangari, S. . Jain, A. . Nilawar, G. . Deshmukh, and A. . Yeole, “Optical Mark Recognition Evaluation System using Dual-Component Approach”, Int J Intell Syst Appl Eng, vol. 12, no. 10s, pp. 349–353, Jan. 2024.
46. Mondal S, De P, Malakar S, Sarkar R. OMRNet: A lightweight deep learning model. Multimed Tools Appl 2024; 83: 14011–14045.
47. Berg A, Borensztein E, Pattillo C. Assessing early warning systems. IMF Staff Pap 2005; 52: 462–502.
48. Stark JT. Enhancing early warning systems with dynamic selection. 2024.
49. Rose TM. Lessons learned in large classroom demonstrations. Am J Pharm Educ 2018; 82: 1081–1085.
50. Wang W. Decision support systems for education management. Comput Intell Neurosci 2021.
51. Hsiao IH, Huang PK, Murphy H. Reviewing and reflecting behaviors from paper-based assessments. In: Proc ACM Int Conf; 2017. pp. 319–328.