The data set is the "Forest Cover Type Dataset" obtained from kaggle. I decided to work on this dataset because it is highly imbalanced, it highlights many different features (some of which are categorical, some of which are continuous), and it involves 7 different classes.
I have looked a new strategy for visualization of the Principle Component Analysis (PCA) instead of two component PCA plots which gives me more insights of the data. This new visualization techniques for analyzing the data is confirmed when we look at the final model.
Here are a list of descriptions for each folder in this reposition
-
data_exploration.ipynb
The file contains the data exploration and visualization techniques, such as PCA.
-
Modeling.ipynb
The file contains the modeling process such as dimension reduction and feature extraction, Random Forest hyper parameter tuning.
For a high level overview in blog post form check out this link: medium