Education is a topic of major importance in today’s societies. It is in our best interest that students achieve high performances, and many times this task is relegated to teachers.They must identify students with difficulties, and ensure their appropriate development. This is commonly done through heuristics or intuition. However, with the number of records available, data can help us make more informed decisions. We can use models to predict student performance, and effectively categorize them into different groups (low, middle, or high level) to more appropriately adapt to their needs, as is done in the Korean Middle School system.
In this project, we aim at understanding the factors that drive student success, and with these insights create a robust model, capable of categorizing students into the mentioned groups. We use EDA to explore our data, its possible problems and what attributes seem to be predictive of success. With these ideas, we do some Feature Engineering, before finally moving on to modelling.
The results are very interesting, and show how the key to success is having well-rounded students, with a nice work-life balance.
For further reading:
Research paper and presentation are available in the repository