Most Machine Learning models treat each row as an independent example to provide a prediction. But what if you want to make a forecast by looking at a series of rows over time? Then you need to know about Time Series!
This mission explores the basic concepts of time series, shows you how to make single series and multiseries models with the AutoTS component of the Datarobot platform. It covers many of the idiosyncrasies and best practices of Time Series modeling through the perspective of several use cases, from simple to more complex.
Format and duration: virtual instructor-led, hands-on, 7 hours
By the end of this mission, you will be able to:
- Understand how time series modeling is different from other Machine Learning
- Set Feature Derivation Windows, Forecast Distances, and Gaps
- Mark features as Known In Advance
- Import custom calendar files of key dates
- Choose the Series ID for multiseries models
- Build and evaluate time series forecast models
- Investigate derived features and the automated feature derivation log
- Utilize a variety of auto-generated feature lists and add your own
- Compare models using familiar and time-series-specific metrics
- Validate time series models with backtesting strategies
- Assess model performance at a specific point, over time, and between series
- Construct a scoring dataset for time series models
- Define the forecast point or range for scoring new data
- Inform your modeling with best practices and advanced techniques like hierarchical modeling, dynamic thresholding, and short series evaluation
Who should complete this mission?
This mission is intended for people who have used DataRobot at an introductory level (at least), and who want to expand their knowledge of the platform’s capabilities.
Before embarking on this mission, you should complete one of the following:
You should have completed either one of these missions, but should not expect to take both.
Technical requirements
- Chrome browser
- The latest version of Zoom
- During class, you will connect to the DataRobot classroom account and download files. If your organization uses security protocols that block connecting to the classroom account or downloading files (for example, a firewall or VPN), please connect to the session using a personal computer.
Completion requirements
Upon completing this mission, you will receive a Certificate of Completion. Attendance of the instructor-led session as well as passing the class assessment are required for mission completion.
Please use this time zone conversion tool to find a session that works best for your schedule and time zone: worldtimebuddy.com
PST/PDT: Americas/Pacific (San Francisco/Seattle)
EST/EDT: Americas/Eastern (Boston/New York/Washington DC)
GMT/BST: Europe/UK (London)
SGT: Asia/Singapore (Singapore)
AEST/AEDT: Australia/Eastern (Sydney/Melbourne)
Important Note
To ensure participants are properly configured and enabled across our internal systems, session registration closes 2 business days prior to the start of the session. Please refer to our Education Services Policies for additional info.