Scikit-Learn offers a handy module for saving and loading trained models:
from sklearn.externals import joblib
To save model:
joblib.dump(model, "model.pkl")
When we split a dataset into test and train sets, we often use different tactics for splitting. Very often methods are based on random records picking and putting them in different sets.