This repository contains code to use various feature selection methods in machine learning models. Lot of times, we have limited data with high dimensionality.
In those cases, it is important to find features which are most important for prediction. This repository provides code for some of the well known feature selection algorithms such as:
- Max-Relevance Min-Redundancy (MRMR)- http://home.penglab.com/papersall/docpdf/2005_TPAMI_FeaSel.pdf
- Split Conformal Inference- https://arxiv.org/pdf/1604.04173.pdf
- Leave one covariate out (LOCO)- https://arxiv.org/pdf/1604.04173.pdf
I will be adding other algorithms as well in the near future.
One of the techniques that I used in my current research is a hybrid combination of mRMR and LOCO.