AI-Augmented Online Aptitude Exam System with Adaptive Testing, Intelligent Question Generation and Performance Analytics

Year : 2026 | Volume : 17 | Issue : 02 | Page : 48 62
By

Shweta Singh,

Riya Maurya,

Mira Gohil,

  1. Research Scholar, Integrated MCA, Thakur Institute of Management Studies, Career Development & Research (TIMSCDR), Maharashtra, India
  2. Research Scholar, Integrated MCA, Thakur Institute of Management Studies, Career Development & Research (TIMSCDR), Maharashtra, India
  3. Mira Gohil, Integrated MCA, Thakur Institute of Management Studies, Career Development & Research (TIMSCDR), Maharashtra, India

Abstract

In the non-technological world, assessment is inflexible, catering for the diverse ability range of test takers which automatically leads to sub-optimal assessment, and sub-optimal engagement with the assessment. We have been developing an open source, AI-based exam system which will be able to meet these challenges. The Item Response Theory (IRT) and standard machine learning procedures are used. The system adjusts the difficulty level of the questions dynamically as per Three-Parameter Logistic (3PL) mode, Maximum Likelihood Estimation or Fisher Information while taking the test. A 2,500 question bank was created and managed by a hybrid artificial intelligence architecture to resolve the issue of manual question calibration. This takes a different strategy: a BERT model is employed to obtain text-based semantic features like text diversity and then the text-based features output by the BERT model is fed into an ensemble Linear Classifier (Naive Bayes, Logistic Regression, and Random Forest). This ensemble method was able to accurately predict the difficulty of the items with 91.7% accuracy. Designed to be scalable, it will consist of a flask-restful API back-end and a NextJS front-end and use AI Fairness 360 toolkit to assist the algorithm in identifying and neutralising bias. It was Experience-tested and validated in 12 weeks live deployment with 247 participants and 847 concurrent users. Outcomes showed a very strong infrastructure with a response time of 187 milliseconds and 99.73% uptime. able It was adaptively shown to improve the accuracy of measurement by 42.5%, and reduce the total number of tests by 40.8% and the test duration by over 37% respectively. In general, such an unambiguous and flexible system inspired students to acquire more knowledge, increased their self-efficacy and was a far more trustworthy, cost-effective replacement for Wealth computerized adaptive assessment environments.

Keywords: Adaptive testing, item response theory, educational assessment, machine learning classifiers, BERT, performance analytics, real-time systems, educational technology

[This article belongs to Journal of Computer Technology & Applications ]

How to cite this article: Shweta Singh, Riya Maurya, Mira Gohil. AI-Augmented Online Aptitude Exam System with Adaptive Testing, Intelligent Question Generation and Performance Analytics. Journal of Computer Technology & Applications. 2026; 17(02):48-62.
How to cite this URL: Shweta Singh, Riya Maurya, Mira Gohil. AI-Augmented Online Aptitude Exam System with Adaptive Testing, Intelligent Question Generation and Performance Analytics. Journal of Computer Technology & Applications. 2026; 17(02):48-62. Available from: https://journals.stmjournals.com/jocta/article=2026/view=260254

References

  1. Wainer H, Dorans NJ, Flaugher R, Green BF, Mislevy RJ. Computerized adaptive testing: A primer. Routledge; 2000 Apr 1.
  2. Hambleton RK, Swaminathan H. Item response theory: Principles and applications. Springer Science & Business Media; 2013 Nov 11.
  3. Devlin J, Chang MW, Lee K, Toutanova K. BERT: Pre-training of deep bidirectional transformers for language understanding. In: Proceedings of NAACL-HLT; 2019. p. 4171-4186.
  4. Breiman L. Random forests. Machine learning. 2001 Oct;45(1):5-32.
  5. Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, Kaiser Ł, Polosukhin I. Attention is all you need. Advances in neural information processing systems. 2017;30.
  6. Domingos P, Pazzani M. On the optimality of the simple Bayesian classifier under zero-one loss. Machine learning. 1997 Nov;29(2):103-30.
  7. Richardson L. Beautiful Soup Documentation — Beautiful Soup 4.4.0 documentation. Crummy.com. 2019. Available from: https://www.crummy.com/software/BeautifulSoup/bs4/doc/ .
  8. Scrapy 2.16 documentation — Scrapy 2.16.0 documentation. Scrapy.org. 2026. Available from: https://docs.scrapy.org/en/latest/ .
  9. Van der Linden WJ, Glas CA, editors. Computerized adaptive testing: Theory and practice. Dordrecht: Kluwer Academic; 2000 Jul 31.
  10. Thissen D, Steinberg L. A taxonomy of item response models. Psychometrika. 1986 Dec;51(4):567-77.
  11. Bellamy RK, Dey K, Hind M, Hoffman SC, Houde S, Kannan K, Lohia P, Martino J, Mehta S, Mojsilovic A, Nagar S. AI Fairness 360: An extensible toolkit for detecting, understanding, and mitigating unwanted algorithmic bias. arXiv preprint arXiv:1810.01943. 2018 Oct 3.
  12. Grinberg M. Flask web development. “ O’Reilly Media, Inc.”; 2018 Mar 5.
  13. Next.js Hosting | Deploy Fast with Hostinger. Hostinger. 2026. Available from: 1 https://www.hostinger.com/in/web-apps-hosting/nextjs-hosting?utm_source=google&utm_med
    ium=cpc&utm_id=23479166145&utm_campaign=Generic-Hosting-Node.js|NT:SE|Lang:EN|
    LO:IN&utm_term=next.js&utm_content=806857457004&gad_source=1&gad_campaignid=23479166145&gbraid=0AAAAADMy-hY_Vi5n_krleaInMMAc1sK12&gclid=Cj0KCQjwjIPSBhCC
    ARIsABGyK7v6-C7L2f1GffvnlYCEyOXdBjoPdOI-VPkaABstJC7teP_MxXo7koEa
    AuLrEALw_wcB.
  14. Scikit-learn: machine learning in Python — scikit-learn 1.9.0 documentation. Scikit-learn.org. 2026. Available from: https://scikit-learn.org/stable/ .
  15. Supabase Docs. Supabase.com. Supabase Docs; 2026. Available from: https://supabase.com/docs.
Regular Issue Subscription Original Research
Volume 17
Issue 02
Received 10/03/2026
Accepted 26/06/2026
Published 30/06/2026
Publication Time 112 Days

Login

My IP

PlumX Metrics

Support