saket-maheshwary/AutoLearn

AutoLearn, a domain independent regression-based feature learning algorithm.

AutoLearn

This repository contains source code for the research work described in our IEEE ICDM 2017 paper :

AutoLearn — Automated Feature Generation and Selection

Dependencies

• Scikit-learn for modelling
• Pandas for data manipulation
• Numpy for performing mathematical operations
• Matplotlib for plotting 2D graphs

How to Run

python main.py

Note

-- The current version of the code is not optimized. I will updated the optimized version in coming weeks.

-- The hyperparameter value of the thresholds are hard coded. Use hyperparameter optimization for best results.

-- All dataset files should follow the exact same template as that of sonar.csv (class labels column at the last)

Citation

If you use our code, please cite the following papers:

@inproceedings{kaul2017autolearn,
title={AutoLearn—Automated Feature Generation and Selection},
author={Kaul, Ambika and Maheshwary, Saket and Pudi, Vikram},
booktitle={Data Mining (ICDM), 2017 IEEE International Conference on},
pages={217--226},
year={2017},
organization={IEEE}
}   


@MISC{saket:automl_pdf,
  AUTHOR = "Maheshwary, Saket and Kaul, Ambika and Pudi, Vikram",
  TITLE = "Data Driven Feature Learning",
  MONTH = Aug,
  YEAR = 2017,
  NOTE =    "\url{https://www.researchgate.net/profile/Saket_Maheshwary/publication/325736313_Data_Driven_Feature_Learning/links/5b20e25ca6fdcc69745d796c/Data-Driven-Feature-Learning.pdf}"
}