[FEATURE] DictMapper

In this example you pass the transformer a column to transform but also a dict. In this case you’d pass;

{'young': 1, 'adult': 2, 'old': 3}

A transformer that transformers based on a dictionary. I don’t know if sklearn has this feature but I couldn’t find it.

[koaning]

@kayhoogland this is why I usually wait until @MBrouns gives feedback before I implement something 😉.

From my end: I’ve tried using the Ordinal Encoder in the past but it is not an intuitive API.

I can imagine if you map strings to numeric values you could surely get some gridsearch value out of it though. I like you could enfore monotonicity across a discrete variable this way and how you might add the hard “up-down-up” shape too if you’d like.

{"young": 1, "adult": 2, "old": 3}
{"young": 1, "adult": 2, "old": 5}

Suppose you’d use these mappings in a disease suseptibility study, I can imagine it can be use-ful to gridsearch.

[MBrouns]

Some things that might be worth discussing:

1. Why does this need to live in a transformer? I don’t expect to gridsearch over it and it doesn’t contain learned properties as far as I can tell.

2. Isn’t this what the OrdinalEncoder in sklearn does?