Showing results for 
Search instead for 
Did you mean: 
Hi Paxata Community members! Welcome to the DataRobot Community! You will find all the Paxata content you know and love— CLICK HERE.

Improving Time Series Models

Community Team
Community Team
0 3 619

(Part of a model building learning session series.)

Time series forecasting is a critical component of any business when solving problems such as demand forecasting, staffing, inventory management, and more. In today's world, leveraging automated machine learning for such use cases is paramount for maintaining a competitive edge. However, due to real-world complexities, additional strategies are often needed to achieve strong performance.

In this learning session, DataRobot's Jess Lin and Taylor Larkin will discuss tips and tricks to improve time series models. Topics will include:

  • Problem framing for optimal results
  • Data preparation for increasing performance
  • Adjusting DataRobot project settings
  • Advanced time series blueprints
  • Best practices for model evaluation


  • Jess Lin (DataRobot, Data Scientist)
  • Taylor Larkin (DataRobot, Data Scientist)
  • Jack Jablonski (DataRobot, AI Success Manager)

Now what?

After watching the learning session, you should check out these resources for more information.

DataRobot Community: 

DataRobot Pathfinder (Use Case Library) — see this query for Time Series use cases


Also, if you have comments or questions that weren't answered in the learning session, you can send email to or click Comment (below) and post them now. We're looking forward to hearing from you!


DataRobot Employee
DataRobot Employee

@taylor_larkin @JessLin How do you deal with monitoring regime shift?

DataRobot Employee
DataRobot Employee

@taylor_larkin @JessLin  You mentioned traditional econometric methods... will you ever cover VAR, VECM, SUR?

Data Scientist
Data Scientist

You mentioned traditional econometric methods... will you ever cover VAR, VECM, SUR?


Currently, we have VAR and VARMAX in the DataRobot platform as well as special versions of these that integrate calendar events and Fourier terms. In my view, the largest advantage of utilizing a VECM or SUR is for inference-related tasks (parameter estimation, significance testing, etc.) when traditional assumptions are violated. While we currently don’t support these methods, if they provide significantly better forecasts for your use cases, we’d love to hear about it! We’re always looking for additions to add to the product, especially if they’re effective across various domains.

Looking for live, instructor-led classes? See details: DataRobot University