It doesn’t take a 50 engineers team and a 1M lines of code project to predict anything these days.
new_pred = regressor.predict(new_dataset)
Just one line of code. As simple as that.
The real work (for which you need brain power more than lines of code) is in:
1) Selecting the variables that truly matter to predict a new variable
2) Compose and clean up the training data set (and keep it updated!)
3) Try different prediction models, score them and use the most efficient
4) Provide a dataset with known values to predict unknown values
5) Format the outputs in a way that makes business sense
For a simple supervised learning / linear regression prediction of reach within Facebook, for a given budget and a given age group, you can check out my -short- example here: https://github.com/leocelis/predictions/blob/master/predict_reach.py