Student Exams Results Prediction

Do you know you can predict the results of a student before they even write the exams using the way they live? I created a model to predict possible results of a student from how ttheylive.

Comment 1

Inspiration

Why we built this Many students believe they can simply study just before exams and still perform well — a misconception that often leads to poor outcomes. This observation prompted us to develop the Student Exam Results Prediction system: a tool that helps students, parents, and teachers estimate academic performance based on a student's everyday habits and behaviours, well before exam day. It also allows educators to calculate how much additional support a student may need to pass.

What it does Product overview The system predicts a student's likely exam outcome by analysing key behavioural factors — including sleep patterns, internet access, and school attendance. Based on this analysis, it classifies each student into one of four performance categories:

Excellent Good Satisfactory Fail

How we built it

Development process The project was built in three stages:

  1. Data preprocessing — We sourced a dataset from Kaggle, then cleaned and prepared it for modelling.
  2. Modelling — We trained and evaluated two models: a Random Forest Classifier (92% accuracy) and a Logistic Regression (23% accuracy). We selected the Random Forest Classifier for its significantly superior performance.
  3. Web app and deployment — We built the front-end using HTML and CSS, and deployed the application via GitHub.

Challenges

What we ran into Our primary technical challenge is deploying the application online via GitHub Pages — the trained model file is too large for standard hosting constraints. Additionally, as full-time students, managing this project alongside our academic workload significantly limited the time available for development.

Accomplishments

What we're proud of The application is hosted on GitHub and is fully functional when run locally via the terminal. Achieving a 92% model accuracy is a result we are particularly proud of, especially given our time constraints.

Learnings

What we learned This project taught us how to handle large machine learning model files within the constraints of web deployment — a practical challenge we had not encountered before.

What's next

Plans We plan to develop a mobile application to make the tool accessible on smartphones, broadening its reach to students, parents, and teachers who primarily use mobile devices.

Built With

Share this project:

Updates