Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM66432
Volume:
10 |
7 |
Jul 2026
Pages:
1-8
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
Computer Science
City Engineering college
Computer Science and Engineering
City Engineering college
Computer Science and Engineering
City Engineering college
Educational institutions continuously seek efficient methods to monitor student performance and provide timely academic guidance. Conventional evaluation methods rely largely on manual analysis of examination scores, which makes it difficult to identify students who require additional academic support before outcomes become irreversible. This paper presents an AI-assisted Student Performance Prediction System that integrates supervised machine learning with a generative-AI academic advisor to forecast academic outcomes and generate individualized recommendations. Unlike earlier systems that analyze only one or two subjects in isolation, the proposed system jointly considers six core Computer Science subjects — Data Structures, Database Management Systems, Operating Systems, Computer Networks, Python Programming, and Artificial Intelligence — together with Reading Score, Writing Score, daily Study Hours, Gender, and Course Preparation Status. Three supervised classifiers, namely Decision Tree, Random Forest, and Logistic Regression, were trained and compared on a dataset of 1000 student records (800 training, 200 testing) to classify students into PASS or FAIL categories. Logistic Regression achieved the highest test accuracy of 99.50%, followed by Random Forest at 97.50% and Decision Tree at 95.50%. Feature-importance analysis further revealed that Study Hours is the single most influential predictor, accounting for approximately 44.3% of the Random Forest model's decision weight. Beyond classification, an AI-powered academic advisor built on the Llama 3.2 large language model, served locally through Ollama, converts the numeric prediction into a personalized performance summary, topic-level improvement suggestions for each weak subject (e.g., specific chapters or concepts to revisit rather than generic advice), and a structured study plan. The platform further incorporates a Teacher's Suggestion module through which an instructor's free-text remarks are captured and merged into the same report alongside the AI-generated content, combining automated analytics with human-in-the-loop academic judgment. The system compiles the prediction, the topic-level guidance, and the teacher's remarks into a downloadable PDF report. The complete pipeline is implemented as a working Streamlit prototype using Python, Scikit-learn, and Ollama, featuring an interactive input form, a prediction-confidence gauge, an academic radar chart, and an automatically generated PDF report. Experimental results confirm that combining interpretable predictive analytics with generative-AI feedback yields both high classification accuracy and actionable, human-readable guidance, making the system practical for adoption by academic institutions and individual educators alike.
KEYWORDS
This paper presented an AI-Assisted Multi-Subject Student Performance Prediction System that combines supervised machine learning with a generative-AI academic advisor to support academic decision-making. By jointly analyzing six core Computer Science subjects together with study habits and preparation status, the evaluated models achieved strong classification performance, with Logistic Regression reaching 99.50% test accuracy and Random Forest reaching 97.50% while additionally providing interpretable feature importance. The integration of a topic-level AI Academic Advisor built on Llama 3.2, together with the Teacher's Suggestion module, extends the system beyond conventional pass/fail prediction by combining automated, curriculum-grounded recommendations with the instructor's own judgment in a single downloadable report. This combination assists educators in identifying at-risk students early and enables timely, individualized academic intervention, making the proposed system a practical and extensible solution for modern educational institutions.
[1] “Student Academic Performance Prediction using Machine Learning with Various Features and Scenarios,” IEEE Conference Publication, 2024. [Online]. Available: https://ieeexplore.ieee.org/document/10770729/ [2] “Student Performance Prediction System Using Machine Learning,” IEEE Conference Publication, 2025. [Online]. Available: https://ieeexplore.ieee.org/document/11118334/ [3] “Predicting Student Performance to Boost Educational Outcomes: The Efficacy of a Random Forest Approach,” IEEE Conference Publication, 2024. [Online]. Available: https://ieeexplore.ieee.org/document/10541035/ [4] “Predicting Students Academic Performance using an Improved Random Forest Classifier,” IEEE Conference Publication, 2020. [Online]. Available: https://ieeexplore.ieee.org/document/9167547/ [5] “Research on Student Performance Prediction Based on Random Forest Algorithm,” in Proc. 2024 Int. Symp. Artificial Intelligence for Education, ACM, 2024, doi: 10.1145/3700297.3700385. [6] “Grade Prediction Based on Student Performance using Gradient Boosting and Random Forest,” IEEE Conference Publication, 2025. [Online]. Available: https://ieeexplore.ieee.org/document/11121505/ [7] M. T. Sathe and A. C. Adamuthe, “Comparative Study of Supervised Algorithms for Prediction of Students’ Performance,” Int. J. Modern Education and Computer Science, vol. 13, no. 1, pp. 1–21, 2021, doi: 10.5815/ijmecs.2021.01.01. [8] “Predicting Students’ Academic Performance Via Machine Learning,” Journal of University of Babylon for Pure and Applied Sciences, vol. 27, no. 1, pp. 194–205. [9] “Machine Learning-Driven Student Performance Prediction for Enhancing Tiered Instruction,” arXiv:2502.03143, 2025. [10] “Using Machine Learning to Predict Student Outcomes for Early Intervention and Formative Assessment,” PMC, 2025. [Online]. Available: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12615791/
Publication Date
7/31/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
1
Total views since publication
PDF Downloads
21
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...