Getting started

Installation

The process is similar to installing other Python packages. The Python version of boruta-fru is available as boruta-fru on PyPI. Use your favorite package manager to install it. The package requires python >= 3.12.

For this tutorial, you will also need scikit-learn, pandas and polars.

Basic usage example

You can initialize a model and use the fit and final_decision functions. The fit function is used to perform feature selection. After running fit, the final_decision function returns a decision for each feature. The model accepts a max_runs parameter during initialization. If there are many Tentative decisions, increasing this parameter may help. Additionally, increasing the number of trees may improve the quality of the decisions.

Warning

Pandas fully supports the PyCapsule interface starting from version 3. We support pandas version 3 and above.

Warning

The target must be a categorical series with string categories or numerical series.

Passing NumPy arrays is not supported. If you are using NumPy arrays, convert them to pandas.DataFrame and pandas.Series.

from sklearn.datasets import load_breast_cancer
from boruta_fru import Boruta

# load data
data = load_breast_cancer(as_frame=True)

# convert target to a categorical series with string categories
y = data["target"].astype(str).astype("category")

# create model instance
boruta = Boruta(max_runs=200, trees=100)

# fit model
boruta.fit(data["data"], y)

# make predictions
boruta.final_decision()

The result of final_decision is an array containing a decision for each column in the original data frame.

• Confirmed means the feature was accepted as important for prediction.

• Rejected means the feature was determined not to be important for prediction.

• Tentative means the feature was neither rejected nor confirmed; therefore, it is unclear whether the feature should be considered important for prediction.

Polars example

Another feature of boruta-fru is that it works with libraries supporting the Arrow PyCapsule interface, such as pandas, polars, pyarrow, duckdb, and others.

import polars as pl

from sklearn.datasets import load_breast_cancer

from boruta_fru import Boruta

# load data
data = load_breast_cancer(as_frame=True)

# create model instance
x = pl.from_pandas(data["data"])
y = pl.from_pandas(data["target"].astype(str).astype("category"))

# create model instance
boruta = Boruta(max_runs=200, trees=100)

# fit model
boruta.fit(x, y)

# make predictions
boruta.final_decision()

Missing values

boruta-fru supports missing values. By default, in each iteration the algorithm imputes missing entries by randomly sampling (with replacement) from the non-null values in the same column.

To disable this behavior, set impute=False.

import pandas as pd

from sklearn.datasets import load_breast_cancer
from boruta_fru import Boruta

# load data
data = load_breast_cancer(as_frame=True)
x = data["data"]
x.iloc[0, 0] = pd.NA

# convert target to a categorical series with string categories
y = data["target"].astype(str).astype("category")

# create model instance
boruta = Boruta(max_runs=200, trees=100)

# fit model
boruta.fit(data["data"], y)

# make predictions
boruta.final_decision()

Progress

The package provides progress reporting via logging. After each iteration, a log message is emitted containing the iteration number and the counts of Tentative, Rejected, and Confirmed features.To enable logging, you just need to change log level to info.

import logging

from sklearn.datasets import load_breast_cancer
from boruta_fru import Boruta

# load data
data = load_breast_cancer(as_frame=True)

# convert target to a categorical series with string categories
y = data["target"].astype(str).astype("category")

# create model instance
boruta = Boruta(max_runs=200, trees=100)

# configure logging
logging.basicConfig()
logging.getLogger().setLevel(logging.INFO)

# fit model
boruta.fit(data["data"], y)

# make predictions
boruta.final_decision()

Result as PyCapsule (optional)

Results can optionally be returned as an Arrow PyCapsule. This allows them to be loaded into any data frame library supporting the Arrow PyCapsule interface.

import pandas as pd

from sklearn.datasets import load_breast_cancer

from boruta_fru import Boruta

# load data
data = load_breast_cancer(as_frame=True)
X = data["data"]
y = data["target"].astype(str).astype("category")

# create model instance
boruta = Boruta(max_runs=200, trees=100)

# fit model
boruta.fit(X, y)

# convert results from PyCapsule
pd.DataFrame.from_arrow(boruta.final_decision(to_pycapsule=True))