Handling Highly Correlated Features

Handling Highly Correlated Features

Degree of relationship between two variables, often given as r (Pearson correlation coefficient - PCC*)*. It can vary between -1, which means perfect inverse correlation, and +1, which means perfect positive correlation. The fastai-tutorial gives an intuitive understanding. It shows why it is so useful to look at the correlations between a feature (column) and the target variable. It becomes clear, that r is very sensitive to Outlier. The outliers can be clearly seen in the plots (→ useful for Outlier Detection).