How to Mitigate Bias in Machine Learning