Practical Data Analysis Cookbook

Sampling the data

Sometimes the dataset that we have is too big to be used to build a model. For practical reasons (so that the estimation of our models does not take forever), it is good to create a stratified sample from the full dataset.

In this recipe, we will read from our MongoDB database and use Python to create a sample.

Getting ready

To execute this recipe, you will need PyMongo, pandas, and NumPy. No other prerequisites are required.

How to do it…

There are two approaches that one can take: either specify the fraction of the original dataset (say, 20%) or specify the number of records one would like to retrieve from the dataset. The following code shows you how to fetch a fraction of the dataset (the data_sampling.py file):

strata_frac = 0.2 ...