Initial Steps in Multiple Linear Regression with Python

Before model building, you first need to perform the test-train split and scale the features.

Scaling of variables is an important step because, as you may have noticed, the variable ‘area’ is on a different scale with respect to all other numerical variables, which take very small values. Also, the categorical variables that you encoded earlier take either 0 or 1 as their values. Hence, it is important to have everything on the same scale for the model to be easily interpretable.

You have seen the two popular rescaling methods- Min-Max scaling and Standardisation (mean=0 and sigma=1). The advantage of Standardisation over the other is that it doesn't compress the data between a particular range as in Min-Max scaling. This is useful, especially if there is are extreme data point (outlier). Now, let's rescale and fit the data.

Now that you have prepared the data and are done with the test-train split, let’s prepare a heat map and take a look at the correlations between the variables.