How to handle time-related features?


I’m trying to build a predictive model for content popularity and some of my features are time-related, for example, number of likes after 5, 10, 15… minutes from publishing. I can build an RNN model but I was wondering if I can build a model with a set of continuous features instead, based on the Taxi challenge on Kaggle:

  1. Use the first 5 data points for each record as 5 features. The measurements interval isn’t accurate so I’ll have to interpolate.
  2. Add 5 data points with their age (10 features in total). This way I don’t have to interpolate and I also don’t have to use the first 5 data points every time.

Option 2 sounds better but I’ve never seen something similar.
I’d appreciate your feedback.