This release is a big one! We've made some major enhancements to all aspects of Kraken. We've added new ways to get data into Kraken, more robust model building, cleaner look and feel of the UI, more detailed information on the models that trained, and much more.
- Submit a Bug: https://bigsquid.zendesk.
com OR Click the big blue box on the Kraken start screen
- Submit Feedback or Feature Requests: https://portal.
prodpad.com/c0a8b130-6d54- 11e7-a1da-22000a2145da OR clic k on the big purple box on the Kraken start screen
- New Data Provider - We've added a new way to get data into Kraken. You can now add a direct connection to MySQL databases. Navigate to the Manage Data screen from the Main Page and enter in your MySQL credentials
- Refresh Data Providers - We've added a couple of ways to refresh the data that come from connected Data Providers. When searching for a Dataset, there is now a link to refresh the datasets coming from your connected Data Providers. This will pick up any newly added Datasets since the time you first connected to the Data Provider. This functionality is also contained in a menu in the top right corner of the Select Dataset screen
- Look and Feel - We've transitioned to a new UI component library, which means the look and feel of Kraken will be more consistent going forward
- Model Training Pipeline - We've made major enhancements to the model training pipeline. We are automatically applying logic to your data to help produce better predictive models. Some of the changes include handling NULL values in your data, better handling of categorical data, and better use of Drivers that are highly correlated to the Predicted Metric or other Drivers
- Deep Dive Analysis Results - We have added more fit statistics for binary classification models to the Deep Dive screen on the Analysis Results page. This will allow you to see even more information on what models were tried by Kraken
- Issues connecting and pushing data to MySQL from a Looker Data Provider
- Various bugs and improvements