Day 67 – Sampling Bias

Sampling bias occurs when A and B compete with each other, the more population of A data is considered. However, the actual outcome B wins.

Take an example, and you are going to see the match between the Seahawks vs. Patriots super bowl match in Seattle, US. When you ask everyone in the stadium who will win the match, obviously, Seahawks fans more in-home stadium and data outcome “Seahawks will win the match.”  If the match happens, Patriots home stadium, then the data outcome “Patriots will win the match.”

If the data is sampled in a biased way, learning will produce biased outcome.

Source: Book “Learning from data”.

We have to make sure training and test data distributions are not biased in any circumstances to get good results.