Random Forest Prediction based on one feature?

(Adnan Sabanovic) #1

I need some help. So I took what I have learned in ML course, lessons 1 through 5 on Random Forest and I wanted to apply to my idea of predicting car prices on the local market website.

I scraped the website and got a few sets ranging from 3k to 8k results for the published cars for sale.

I have about 21 features including (Manufacturer, Model, Made, Volume …etc)

Now If I am about to put this in production and actually make the prediction about the prices, I am guessing I shouldn’t ask my users to type every single feature(Keep in mind that I have ‘date’ feature that got encoded into a dozen of those).

Is there a way to to m.predict(inputs) where inputs is a an array containing only Manufacturer and Model ?

I am pretty sure this is a well known scenario so would appreciate any response.

Thank you

(Krishna Kishor Kammaje) #2

Did you think about passing the remaining columns as median/mid-value (for categorical).?

(Adnan Sabanovic) #3

I actually have and I think that makes sense, with a note for the user that selecting only one feature would give lower precision.

Thank you