Ambiguities regarding data selections for features

Good evening.
While selecting features for a particular model, there may be some features that can be quantified, while some features are binary or with a few classes. Can we choose a model for ANN based on some features quantifiable as well as features that are non quantifiable?
For example, if am interested in calculating the overhead cost of construction project, I may choose features like inflation which can be quantified, or features like field safety measures which can either be present or absent. Similarly extreme weather like rain or snow can halt the project while high temperature may causes delays. There are may be many other features. My question is how to I transform this data before it can be entered as in input for the model?

Yes, you can have mixed types of input features.

  • Any categorical inputs are converted to one-hot representation. So if you have a feature that has ā€˜Nā€™ labels, that turns into ā€˜Nā€™ binary inputs, where N-1 of them are set to 0, and one is set to 1.
  • Categorical features are not normalized.
  • Real-valued inputs may be included, but they should be normalized.