Time Series with Zillow’s Luminaire — Part III Modeling

Time series modeling connects the dots in the starry night; time series forecasting extends the dreams as the breeze brushes the stars into trails. You and I, standing in awe, look at the blue summer night.

In Part I and II of the series “Time Series with Zillow’s Luminaire”, I have walked you through the data exploration and model specification Steps, now we are ready for modeling!

The luminaire offers two main approaches for time series modeling: (A) Kalman Filter modeling and (B) Structural modeling. If this is the first time you've heard of the Kalman Filter, you may not know that numerous devices in our lives have relied on it. The Kalman Filter estimates the trajectory of a moving object. Your iPhone or Android phone has a map app that estimates the location of the phone and driving distance. Cars, fleet trucks, ships, aircraft, or drones have the GPS (Global Positioning System) to track movement with more accuracy. A famous early use was the Apollo navigation computer that took Neil Armstrong to the moon and, most importantly, brought him back.

Further, a time series can be considered as the combination of patterns such as a linear trend, seasonality, and holiday effect. So structural modeling (B) tried to model these patterns explicitly.