Check correlations between features

Correlation Matrix is an useful tool which provides an overview of correlations between features.

Note that categorical features are greyed out.

In the Correlation Matrix tab, you can see a N-by-N grid of features where N is the number of features in your dataset.

• Positive correlation is indicated in warm colours, which approximates red as it approaches 1
• Negative correlation is indicated in cold colours, which approximates purple as it approaches -1
• When there is no strong correlation between two features (around 0), the colour gets closer to white

Positive Correlations

With the Iris dataset, for example, you can easily see that the PetalLengthCm and SepalLengthCm features are positively correlated, meaning that the larger the sepal, the larger the petal.

Negative Correlations

On the other hand, with the world-dataset-2023.csv dataset, you can see that infant_mortality_rate and life_exp_at_birth has a negative correlation of -0.49, meaning that the higher the infant mortality rate, the lower the life expectancy.

Removing colinear features

Correlation matrix can be useful for removing features that are perfectly/extremly highly correlated.

For instance, in the world-dataset-2023.csv, there are two measurements called birth rate and fertility rate, with a correlation of 0.98. To simply the linear regression model, one can consider dropping one over the other.