A User-Guided Bayesian Framework for Ensemble Feature Selection
Admissibility for constraint group
Build a customized constraint for UBayFS
Build an ensemble for UBayFS
Perform stratified data partition.
Build a constraint system
Build decorrelation constraints
Evaluate a feature set
Checks whether a list object implements proper UBayFS user constraints
Check whether an object is a UBaymodel
Posterior expectation of features
Prints the UBayconstraint object
Print a UBayFS model
Run an interactive Shiny app for demonstration
Initial feature set sampling using probabilistic Greedy algorithm
Set constraints in UBaymodel object
Set optimization parameters in a UBaymodel object
Set weights in UBaymodel object
UBayFS feature selection
The framework proposed in Jenul et al., (2022) <doi:10.1007/s10994-022-06221-9>, together with an interactive Shiny dashboard. 'UBayFS' is an ensemble feature selection technique embedded in a Bayesian statistical framework. The method combines data and user knowledge, where the first is extracted via data-driven ensemble feature selection. The user can control the feature selection by assigning prior weights to features and penalizing specific feature combinations. 'UBayFS' can be used for common feature selection as well as block feature selection.
Useful links