Temporal aspects for Linear Regression and other ML Methods

by thebluephantom   Last Updated April 16, 2018 11:19 AM - source

Having completed an interesting course on Linear Regression on coursera, one aspect was never addressed that I am curious about in terms of best approach.

With reference to the classic house price prediction use case:

  • It strikes me that house prices changes over time.
  • What is the best approach therefore?

  • Do we keep on updating the input replacing older house prices of yesteryear?

  • Or do we add an extra feature for Date of Sale - include a temporal aspect as a feature? I mention this as it was never discussed, but at the same time it seems so obvious. Inclusion of such an attrbue means it may get a weight, etc.

I can think of a few things but am more interested in the expert opinion.

Related Questions

Neural Net vs Multiple Linear regression?

Updated March 02, 2017 16:19 PM

Machine learning with multiple independent variables

Updated November 30, 2017 02:19 AM