Project Overview
This machine learning project aimed at exploring and predicting factors that significantly influence students' academic performance. I used different ML algorithms, and aimed to find the most well performing one, and used these models to find the most important features for it.
Features
- Clustering
- Predictive Modeling
- Feature analysis
Challenges & Solutions
The first challenge, as with most machine learning projects, was the data, since it wasn't clean. I preprocessed it, and went on to clustering, which was also quite tough to interpret, since the clusters had revealed a weird way that the grades were formatted
The models, chosing what models to compare, and how, was another difficult challenge. I settled on comparing the r-score, MSE, and MAE. The most difficult thing with the models, was that there were a lot of columns that were correlated with one another, which introduced multicolinearity. This was tough to weed out, as I had to run VIF, to find the features that were highly correlated, and remove them. This is the reason why before running the training on the models, I have removed so many columns
Live Demo & Source Code
View Live Project | Source Code
