EFS: an ensemble feature selection tool implemented as R-package and web-application.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Feature selection methods aim to identify optimal feature subsets for classification models, enhancing prediction performance and interpretability.
• Single feature selection methods often exhibit biases, which can be mitigated by ensemble approaches.

Purpose of the Study:

• To introduce Ensemble Feature Selection (EFS) software for improved feature selection.
• To leverage ensemble methods to compensate for individual feature selection technique biases.

Main Methods:

• EFS integrates eight distinct feature selection algorithms.
• It combines normalized outputs from multiple methods to generate a quantitative ensemble importance score.
• The software can utilize methods individually or in ensemble configurations.

Main Results:

• EFS identifies relevant features by compensating for biases inherent in single methods.
• The ensemble approach leads to improved prediction accuracy in binary classification models.
• Enhanced interpretability of classification models is achieved through EFS.

Conclusions:

• Ensemble Feature Selection (EFS) offers a robust strategy for identifying relevant features.
• EFS enhances the performance and interpretability of binary classification models.
• The EFS software is available as an R-package and via a web application.