We've been working hard to make Kraken more useful and easy to use. This release comes with several exciting new features as well as improvements to exiting features. The all new Help Center will help provide more information on Data Science concepts as you work through building a model. The Scenarios tool got a major face lift to make it even more useful and fun to play with.
TL; DR
Major improvements to Scenarios, time series, and a new Help Center
*** NEW ***
- Kraken Help Center: We've added a new Help Center in Kraken to provide access to documentation and definitions of key terms around Kraken. You can click on the ? icon in the top right and the Help Center will show content that is relevant to the page you are on. It also contains links to our Knowledge Base, Feedback Portal, and Customer Support
- Scenario Tool v2: The Scenarios Tool has gone through some major changes to make it faster and even more useful. All scenario options are shown on a single page and you can quickly iterate through different scenarios to see how they might impact your predictions. Scenarios can be named, you can add notes, and delete any scenarios that aren't needed
*** IMPROVEMENTS ***
- Time Series Optimization: Several optimizations have been made to time series analysis to make it run several times faster. The workflow remains the same, but there is less time waiting on results!
- Correlations on Analysis Results: Once an analysis is complete you can click on the new Correlations button to view the correlation to target and correlation matrix. This will help identify changes you could make to the analysis to produce better results.
*** FIXED ***
- Fixed an issue where special characters in Predicted Dataset names was causing the pushback to Qlik to fail
- Removed the option to apply predictions to a different dataset to remove confusion around what data can be used to create a Predicted Dataset
- Added a fix to better handle null values being sent as a dash (-) from Qlik
- Fixed a UI bug in Deep Dive mode that made it look like a large amount of duplicate models were being trained
- Various other bug fixes and improvement