# Feature Feature Correlation#

This notebook provides an overview for using and understanding the feature-feature correlation check.

This check computes the pairwise correlations between the features, potentially spotting pairs of features that are highly correlated.

Structure:

## How are The Correlations Calculated?#

This check works with 2 types of features: categorical and numerical, and uses a different method to calculate the correlation for each combination of feature types:

1. numerical-numerical: Pearson’s correlation coefficient

2. numerical-categorical: Correlation ratio

3. categorical-categorical: Symmetric Theil’s U

## Imports#

import pandas as pd
from deepchecks.tabular.checks.data_integrity import FeatureFeatureCorrelation


We load the Adult dataset, a dataset based on the 1994 US Census containing both numerical and categorical features.

ds = adult.load_data(as_train_test=False)


## Run the Check#

check = FeatureFeatureCorrelation()
check.run(ds)

# To display the results in an IDE like PyCharm, you can use the following code:
# check.run(ds).show()
# The result will be displayed in a new window.

Feature-Feature Correlation

## Define a Condition#

Now we will define a condition on the maximum number of pairs that are correlated above a certain threshold. In this example, we will define a condition that the maximum number of pairs that are correlated above 0.8 is less than 3.

check = FeatureFeatureCorrelation()