/mrmr

mRMR (minimum-Redundancy-Maximum-Relevance) for automatic feature selection at scale.

Primary LanguagePythonMIT LicenseMIT

drawing

What is mRMR

mRMR, which stands for "minimum Redundancy - Maximum Relevance", is a feature selection algorithm.

Why is it unique

The peculiarity of mRMR is that it is a minimal-optimal feature selection algorithm.
This means it is designed to find the smallest relevant subset of features for a given Machine Learning task.

Selecting the minimum number of useful features is desirable for many reasons:

  • memory consumption,
  • time required,
  • performance,
  • explainability of results.

This is why a minimal-optimal method such as mrmr is often preferable.

On the contrary, the majority of other methods (for instance, Boruta or Positive-Feature-Importance) are classified as all-relevant, since they identify all the features that have some kind of relationship with the target variable.

When to use mRMR

Due to its efficiency, mRMR is ideal for practical ML applications, where it is necessary to perform feature selection frequently and automatically, in a relatively small amount of time.

For instance, in 2019, Uber engineers published a paper describing how they implemented mRMR in their marketing machine learning platform Maximum Relevance and Minimum Redundancy Feature Selection Methods for a Marketing Machine Learning Platform.

How to install this package

You can install this package in your environment via pip:

pip install mrmr_selection

And then import it in Python through:

import mrmr

How to use this package

This package is designed to do mRMR selection through different tools, depending on your needs and constraints.

Currently, the following tools are supported (others will be added):

  • Pandas
  • Polars
  • Spark
  • Google BigQuery

The package has a module for each supported tool. Each module has at least these 2 functions:

  • mrmr_classif, for feature selection when the target variable is categorical (binary or multiclass).
  • mrmr_regression, for feature selection when the target variable is numeric.

Let's see some examples.

1. Pandas example

You have a Pandas DataFrame (X) and a Series which is your target variable (y). You want to select the best K features to make predictions on y.

# create some pandas data
import pandas as pd
from sklearn.datasets import make_classification
X, y = make_classification(n_samples = 1000, n_features = 50, n_informative = 10, n_redundant = 40)
X = pd.DataFrame(X)
y = pd.Series(y)

# select top 10 features using mRMR
from mrmr import mrmr_classif
selected_features = mrmr_classif(X=X, y=y, K=10)

Note: the output of mrmr_classif is a list containing K selected features. This is a ranking, therefore, if you want to make a further selection, take the first elements of this list.

2. Polars example

# create some polars data
import polars
data = [(1.0, 1.0, 1.0, 7.0, 1.5, -2.3), 
        (2.0, None, 2.0, 7.0, 8.5, 6.7), 
        (2.0, None, 3.0, 7.0, -2.3, 4.4),
        (3.0, 4.0, 3.0, 7.0, 0.0, 0.0),
        (4.0, 5.0, 4.0, 7.0, 12.1, -5.2)]
columns = ["target", "some_null", "feature", "constant", "other_feature", "another_feature"]
df_polars = polars.DataFrame(data=data, schema=columns)

# select top 2 features using mRMR
import mrmr
selected_features = mrmr.polars.mrmr_regression(df=df_polars, target_column="target", K=2)

3. Spark example

# create some spark data
import pyspark
session = pyspark.sql.SparkSession(pyspark.context.SparkContext())
data = [(1.0, 1.0, 1.0, 7.0, 1.5, -2.3), 
        (2.0, float('NaN'), 2.0, 7.0, 8.5, 6.7), 
        (2.0, float('NaN'), 3.0, 7.0, -2.3, 4.4),
        (3.0, 4.0, 3.0, 7.0, 0.0, 0.0),
        (4.0, 5.0, 4.0, 7.0, 12.1, -5.2)]
columns = ["target", "some_null", "feature", "constant", "other_feature", "another_feature"]
df_spark = session.createDataFrame(data=data, schema=columns)

# select top 2 features using mRMR
import mrmr
selected_features = mrmr.spark.mrmr_regression(df=df_spark, target_column="target", K=2)

4. Google BigQuery example

# initialize BigQuery client
from google.cloud.bigquery import Client
bq_client = Client(credentials=your_credentials)

# select top 20 features using mRMR
import mrmr
selected_features = mrmr.bigquery.mrmr_regression(
    bq_client=bq_client,
    table_id='bigquery-public-data.covid19_open_data.covid19_open_data',
    target_column='new_deceased',
    K=20
)

Reference

For an easy-going introduction to mRMR, read my article on Towards Data Science: “MRMR” Explained Exactly How You Wished Someone Explained to You.

Also, this article describes an example of mRMR used on the world famous MNIST dataset: Feature Selection: How To Throw Away 95% of Your Data and Get 95% Accuracy.

mRMR was born in 2003, this is the original paper: Minimum Redundancy Feature Selection From Microarray Gene Expression Data.

Since then, it has been used in many practical applications, due to its simplicity and effectiveness. For instance, in 2019, Uber engineers published a paper describing how they implemented MRMR in their marketing machine learning platform Maximum Relevance and Minimum Redundancy Feature Selection Methods for a Marketing Machine Learning Platform.