Introduction

Featurebox contains some tools (Generation and Selection) for material features. Generation is used for feature generation in batch model. Selection is used for feature selection.

And one Extractor in command line mode is add to obtain some special properties in batch model. The special properties need certain subsequent computational processing or third-party software participation.

In total, Batching is the central idea of this module. All works are for convenient data manipulation.

Generation tools

Name	Application
featurebox.featurizers.atom.mapper	atom Getting each element data of compound.
featurebox.featurizers.envir	bond Getting local environment data of compound.
featurebox.featurizers.state	state Getting holistic compound data.
featurebox.featurizers.bond.expander	Tools to transforming pure bond data.
featurebox.featurizers.batch_feature	A built-in goofy tool for generating features.
featurebox.data.namesplit.NameSplit	Dividing compound names to elemental proportion table.
featurebox.data.mp_access.MpAccess	Getting data from pymatgen conveniently.

All the Generation tools with convert method for single case. and fit_transform methods for case list.

Guide: Data Type for Generation

Binding selection tools

Name	Application
featurebox.selection.backforward.BackForward	Backforward selection
featurebox.selection.corr.Corr	Correlation selection.
featurebox.selection.exhaustion.Exhaustion	Exhaustion selection.
featurebox.selection.ga.GA	Genetic algorithm selection.

All the Selection tools are sklearn-type, with fit, fit_transform methods .etc.

Note

Where the binding means treat the binding features as one feature. And the binding features are selected or deleted synchronously.

Guide: Binding Selection

Property batching extractor

Name	Application
featurebox.cli.vasp_bader	Bader Charge
featurebox.cli.vasp_cohp	COHP
featurebox.cli.vasp_dbc	band center
featurebox.cli.vasp_dos	DOS for plot
featurebox.cli	More …

All the Extractor tools with convert method for single case. and fit_transform methods for case list.

Guide: Command Mode for Extractor

Note

The properties batching extractor are suggested to use Command line mode . But interactive model is still available for more customized operation.

Note

The Graph neural network have been removed to pyg_extension package, which employ envir, bond and atom .etc to build input data.