FSelectorRcpp: Entropy-Based Feature Selection Algorithms with Sparse Matrix Support

FSelectorRcpp is an R package implementing entropy-based feature selection algorithms. The implementation is based on the MDL discretization method described by Fayyad and Irani in 1993 and includes support for sparse matrices. The package author is Zygmunt Zawadzki.

Use Cases

• Selecting relevant features for classification models based on entropy measures.
• Reducing dimensionality of datasets with sparse matrix representations.
• Implementing feature selection pipelines using the MDL discretization method mentioned in the description.

Strengths

• Implementation is based on a cited, peer-reviewed algorithm from IJCAI 1993.
• Includes support for sparse matrices, which can improve efficiency for certain data types.
• Provides an Rcpp implementation, which is free of Java/Weka dependencies.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Zygmunt Zawadzki

Collection Method

Software package implementing published algorithms.