Define new features to incorparate into my model

[ricardoj]

I was wondering if someone could give some ideas on how I can incorporate new features on my actual consumer credit score model, so have been testing many different features and combinations, but it is not working maybe I'm doing something wrong; any thoughts would be very appreciated, I'm still learning how to use the platform, and I'm not a data scientist

[ricardoj]

THANK YOU SO MUCHHHHH!!

[Erica · Accepted solution]

I've found the example here: https://community.datarobot.com/t5/ai-ml-knowledge-base/use-case-predicting-loan-defaults/ta-p/740

Might give you some inspiration!

Erica

[Erica]

No problem - how big is your dataset?

There is actually a lot you can add to credit scoring models, as many things can influence it. The data robot demos and examples come with a loans dataset from lendingclub.com.

If you have access to that, that's a real world example that can give you ideas of variables that you can take from your own sources to analyse credit risk - ask your contact at DataRobot if not.

Erica

[ricardoj]

Erika thank you so much for your email, very helpful  I have done that I believe the issue I'm facing is related to my data set all the features I have added are not predicting maybe because I don't have enough data. Thank you 

[Erica · Accepted solution]

Hi Ricardo, could you provide an example?

Out-of-the box DataRobot allows for some custom features to be added to the dataset, based on the existing data.

You can find this functionality in the data screen under Actions sub menu under the main menu - called "Created f(x) transform"

This will allow you to enter a very basic calculation based on one or more fields and add that to your model

Anything more complex, for example joining another dataset or applying a specific filter would need to be done either before you load the data to datarobot, or in Paxata, datarobot's integrated data prep tool.

Erica