lizhen0909/spark-infotheoretic-feature-selection

This package contains a generic implementation of greedy Information Theoretic Feature Selection (FS) methods. The implementation is based on the common theoretic framework presented by Gavin Brown. Implementations of mRMR, InfoGain, JMI and other commonly used FS filters are provided.

An Information Theoretic Feature Selection Framework

The present framework implements Feature Selection (FS) on Spark for its application on Big Data problems. This package contains a generic implementation of greedy Information Theoretic Feature Selection methods. The implementation is based on the common theoretic framework presented in [1]. Implementations of mRMR, InfoGain, JMI and other commonly used FS filters are provided. In addition, the framework can be extended with other criteria provided by the user as long as the process complies with the framework proposed in [1].

Spark package: http://spark-packages.org/package/sramirez/spark-infotheoretic-feature-selection

Please cite as: S. Ramírez-Gallego; H. Mouriño-Talín; D. Martínez-Rego; V. Bolón-Canedo; J. M. Benítez; A. Alonso-Betanzos; F. Herrera, "An Information Theory-Based Feature Selection Framework for Big Data Under Apache Spark," in IEEE Transactions on Systems, Man, and Cybernetics: Systems, in press, pp.1-13, doi: 10.1109/TSMC.2017.2670926 URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7970198&isnumber=6376248

Main features:

• Version for new ml library.
• Support for sparse data and high-dimensional datasets (millions of features).
• Improved performance (less than 1 minute per iteration for datasets like ECBDL14 and kddb with 400 cores).

This work has associated two submitted contributions to international journals which will be attached to this request as soon as they are accepted. This software has been proved with two large real-world datasets such as:

• A dataset selected for the GECCO-2014 in Vancouver, July 13th, 2014 competition, which comes from the Protein Structure Prediction field (http://cruncher.ncl.ac.uk/bdcomp/). We have created a oversampling version of this dataset with 64 million instances, 631 attributes, 2 classes.
• kddb dataset: http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary.html#kdd2010%20%28bridge%20to%20algebra%29. 20M instances and almost 30M of attributes.

Example (ml):

import org.apache.spark.ml.feature._
val selector = new InfoThSelector()
	.setSelectCriterion("mrmr")
      	.setNPartitions(100)
      	.setNumTopFeatures(10)
      	.setFeaturesCol("features")
      	.setLabelCol("class")
      	.setOutputCol("selectedFeatures")

val result = selector.fit(df).transform(df)

Example (MLLIB):

import org.apache.spark.mllib.feature._
val criterion = new InfoThCriterionFactory("mrmr")
val nToSelect = 100
val nPartitions = 100

println("*** FS criterion: " + criterion.getCriterion.toString)
println("*** Number of features to select: " + nToSelect)
println("*** Number of partitions: " + nPartitions)

val featureSelector = new InfoThSelector(criterion, nToSelect, nPartitions).fit(data)

val reduced = data.map(i => LabeledPoint(i.label, featureSelector.transform(i.features)))
reduced.first()

Design doc: https://docs.google.com/document/d/1HOaPL_HJzTbL2tVdzbTjhr5wxVvPe9e-23S7rc2VcsY/edit?usp=sharing

Prerequisites:

LabeledPoint data must be discretized as integer values in double representation with a maximum of 256 distinct values. By doing so, data can be transformed to byte type directly, making the selection process much more efficient.

Contributors

• Sergio Ramírez-Gallego ([email protected]) (main contributor and maintainer).
• Héctor Mouriño-Talín ([email protected])
• David Martínez-Rego ([email protected])

References

[1] Brown, G., Pocock, A., Zhao, M. J., & Luján, M. (2012). "Conditional likelihood maximisation: a unifying framework for information theoretic feature selection." The Journal of Machine Learning Research, 13(1), 27-66.