How to predict time series data using fastai api

(jlam) #1

Hi: I applied the steps in lesson 4, structured data to the New Car Sales dataset in Kaggle and got a pretty low validation loss, see below. But how do I use the model to predict the future sales? For example, let say my dataset ended yesterday, and I wanted to predict sales for the next 3 months, what function should I called from model to generate the sales forecast? Thank you in advanced for your help.

epoch trn_loss val_loss exp_rmspe
0 0.296747 0.192063 0.379568
1 0.148141 0.094192 0.287121
2 0.089151 0.099709 0.278661




Stumpled upon your post when trying to figure out the same issue. Was just wondering if you happened to resolve the problem of getting predictions for future dates?