Related Experiment Video
Updated: Jun 9, 2026
07:35
Selecting Multiple Biomarker Subsets with Similarly Effective Binary Classification Performances
Published on: October 11, 2018
Discriminative and informative features for biomolecular text mining with ensemble feature selection
Sofie Van Landeghem1, Thomas Abeel, Yvan Saeys
1Department of Plant Systems Biology, VIB, Ghent University, Gent, Belgium.
Bioinformatics (Oxford, England)
|September 9, 2010
Summary
Feature selection (FS) enhances biomolecular text mining by identifying key features, improving classifier accuracy and interpretability. This method helps understand predictions and refine text mining tools.
Area of Science:
- Biomolecular text mining
- Machine learning interpretability
Background:
- Black box models in biomolecular text mining hinder prediction understanding.
- Feature selection (FS) offers a solution by identifying relevant features.
Purpose of the Study:
- To apply FS for improved accuracy and interpretability in biomolecular text mining.
- To gain insights into classification predictions for better tool development.
Main Methods:
- Utilized feature selection (FS) methodologies.
- Applied FS to a biomolecular event extraction framework.
- Leveraged Java-ML for algorithms and publicly available datasets.
Main Results:
- FS successfully reduced machine-generated features, boosting classifier performance.
- Identified discriminative features reflecting biological and linguistic patterns.
- Gained insights for enhancing current text mining tools.
Conclusions:
- FS is crucial for improving biomolecular text mining accuracy and interpretability.
- The methodology provides valuable insights into prediction mechanisms.
- This approach facilitates the development of more effective text mining tools.
