I am pre-processing a real-world dataset with some features with missing values. these features are not mandatory - the user doesn't have to provide them. This means the values are not missing at random. Some of the features are nominal while others are contineous.
for example, i'm looking to describe our suppliers by expense, maufecturing location and quality score. By definition, the quality score is not registered to ALL suppliers, since we conduct audits and other assessment to ~30 of our biggest suppliers only.
I'm trying to build a ML prediction algorithm than will utilize this "extra feature" quality score.
Can anyone suggest how to tackle a database composed of un-obligatory features? Is there a model that accepts these kind of features or I have to deal with it in pre-processing stage?
I would most appiciate if you can also refer me to articles/ posts you think are relevant.