The feature selection algorithms MultiSURF (Urbanowicz et al., 2017), MultiSURF* (Granizo-Mackenzie & Moore, 2013), ReliefF (Kononenko et al., 1994), TuRF (Moore & White, 2007), SURF (Greene et al., 2009), SURF* (Greene et al., 2010) are imple- mented as described in Urbanowicz et al. (2017) paper. These feature selection al- gorithms differ in the number of nearest neighbours, algorithms’ computational ef- ficiency, and scoring methodology for selecting near or far instances. Another ap- proach of feature selection involves iteratively removing or adding features to con- struct a feature subset, guided by an estimator. The HSFSI framework provides two meta-transformers implemented using the scikit-learn library based on the impor- tance weights of linear support vector classifier with L1 penalty and extra Trees Clas- sifier with 50 estimators for selecting features
This work is part of Thesis of Chandravesh chaudhari, Doctoral candidate at CHRIST (Deemed to be University), Bangalore, India under supervision of Dr. Geetanjali purswani.
- replicable
- customisable
- Saves time
This project is available at PyPI. For help in installation check instructions
python3 -m pip install feature-selectionpy - Web based GUI