How handle the add of a new feature to the dataset?

Question

Let's say that you trained a model (eg. Random Forest) on a dataset with ten features (or columns).

Now you add one or more features to the dataset. You need the information brought from these new features so you want to enrich your model.

What happens to the data collected till now? How can I use it? Those data have just ten features. How can I use them to train the new model?

You could use just toss them all into the same model with feature hashing, or you could use the older data for representation learning; cf. semi-supervised learning. Once you have enough new data, you could keep it simple and discard the old data. — Emre, Jul 2, 2018 at 16:28
The easiest solution would be to just retrain a model. But, i mean, it also depends on the scenario. Why was this variable not included in the beginning? — Jon, Jul 2, 2018 at 22:55

pcko1 · Accepted Answer · 2018-07-02 18:03:30Z

2

What you ask for is known as Transfer Learning in the Machine Learning framework, so you might want to look more into that direction.

An interesting publication regarding Transfer Learning in Decision Trees is this.

answered Jul 2, 2018 at 18:03

pcko1

3,9201 gold badge14 silver badges29 bronze badges

Add a comment |

1 Answer 1