Data bias is a machine learning term that describes when certain parts of data are weighted more than others in the algorithm or if the dataset does not fully represent the use case for which the algorithm is being trained.
For example:
You want to use machine learning to forecast future customer purchases. But the dataset you use to train the algorithm contains only loyalty customers that purchase frequently.
Using the algorithm on datasets containing customers that purchase infrequently will lead to inaccurate forecasts because the data used to train the algorithm was biased towards frequent purchasers.