Detecting data drift using the Kolmogorov-Smirnov test

After successfully deploying your heart disease prediction model, you've been monitoring its performance and input data. You've noticed that the distribution of some key features in the recent data collected in February looks a bit different from the data you trained on in January. Such discrepancies can affect the model's performance, and it's crucial to detect and address them.

In this exercise, you will use the Kolmogorov-Smirnov (K-S) test to detect any potential data drift between the January and February datasets. Sample datasets called january_data and february_data are already loaded for you.

This exercise is part of the course

End-to-End Machine Learning

View Course

Exercise instructions

Import the ks_2samp function from the scipy.stats module.
Use the provided sample datasets january_data and february_data to perform the Kolmogorov-Smirnov test; calculate the test statistic and p-value.
Check if the p-value is less than 0.05, indicating data drift; if data drift is detected, print "Data drift detected.", otherwise, print "No data drift detected."

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

# Import the ks_2samp function
from ____.____ import ____

# Calculate the test statistic and p value
test_statistic, p_value = ____(____, ____)

# Check the p-value and print the detection result
if ____:
	print("Data drift detected.")
else:
	print("No data drift detected.")

Edit and Run Code