Scaling New Data using Scaler.Transform
new_data_scaled = scaler.transform(new_data)
which mean and std deviation will be used to transform the new_data?
- Mean and Std Deviation from scaler.fit(x) at the beginning of this exercise?
Thank you for reaching out!
That is exactly right. At the beginning of the notebook, we use our "training" data,
data[['SAT','Rand 1,2,3']], to calculate the mean and the standard deviation of each of the features. The result is as follows (the standard deviation is the square root of the variance):
Towards the end of the notebook, when making predictions, we use the same
scaler object (storing the same mean and variance values) to predict the outcome:
In fact, it is very important to use the
StandardScaler() object storing the mean and variance of the training data. The mean and variance of the test data, which in our case is
[1700, 2] and
[1800, 1], should remain unknown to the
scaler object. That is to say, we use the mean and variance of the training data to transform the test data.
Hope this helps!