Siam

As a mathematics student, I wanted to explore a practical question: Can historical exam papers help students understand what they should prioritize before an exam? So I built a full-stack system that analyzes 10 years of university exam papers (2015–2024) and turns them into a student-friendly study guide. What I built: 📚 Syllabus Taxonomy Organizes subjects into Chapters → Subchapters → Topics. 🔎 Recurring Question Detection Identifies questions that represent the same underlying exam problem, even when the wording changes. 📊 Important Chapter Ranking Analyzes historical question patterns and ranks chapters by study priority. 🤖 2025 Question Prediction Uses historical patterns and statistical features to estimate which questions/topics are more likely to appear. 🧪 Walk-Forward Backtesting Instead of randomly splitting the data, the system simulates real exam prediction: 2020 → trained on 2015–2019 2021 → trained on 2015–2020 … 2025 → trained on 2015–2024 📝 Must-Practice Questions Highlights high-priority Section C questions for focused preparation. 💻 Student-Focused Web App Includes search, interactive charts, syllabus exploration, recurring-question history, prediction details, bookmarks, responsive design, and dark mode. The project combines Python, data analysis, embeddings, LLM-assisted semantic judgment, statistical modeling, and frontend development into one workflow. Most importantly, I tried to keep the system honest: the predictions are statistical estimates based on historical patterns—not guarantees of what will appear on an exam. This project started as an idea about finding patterns in past papers and evolved into a complete end-to-end data product. Next step: deploying it publicly and continuing to improve the system.

This is the first raw data for fortran - mathematics 2nd year NU