Normalizing and standardizing the features
We normalize (or standardize) data for computational efficiency and so we do not exceed the computer's limits. It is also advised to do so if we want to explore relationships between variables in a model.
Tip
Computers have limits: there is an upper bound to how big an integer value can be (although, on 64-bit machines, this is, for now, no longer an issue) and how good a precision can be for floating-point values.
Normalization transforms all the observations so that all their values fall between 0 and 1 (inclusive). Standardization shifts the distribution so that the mean of the resultant values is 0 and standard deviation equals 1.
Getting ready
To execute this recipe, you will need the pandas module.
No other prerequisites are required.
How to do it…
To perform normalization and standardization, we define two helper functions (the data_standardize.py file):
def normalize(col):
'''
Normalize column
'''
return (col - col.min()) / (col.max() - col.min())
def standardize(col):
'''
Standardize column
'''
return (col - col.mean()) / col.std()How it works…
To normalize a set of observations, that is, to make each and every single one of them to be between 0 and 1, we subtract the minimum value from each observation and divide it by the range of the sample. The range in statistics is defined as a difference between the maximum and minimum value in the sample. Our normalize(...) method does exactly as described previously: it takes a set of values, subtracts the minimum from each observation, and divides it by the range.
Standardization works in a similar way: it subtracts the mean from each observation and divides the result by the standard deviation of the sample. This way, the resulting sample has a mean equal to 0 and standard deviation equal to 1. Our
standardize(...) method performs these steps for us:
csv_read['n_price_mean'] = normalize(csv_read['price_mean']) csv_read['s_price_mean'] = standardize(csv_read['price_mean'])
