Qlik AutoML Tips - Tip #2 - The data you should be using - (must contain outcomes)

[Michael_Tarallo]

Can't see the video? Watch on the Qlik video site here.

Missed Tip #1? - View it here.

Tip #2  covers the data you should be using with Qlik AutoML. When creating new ML experiments with Qlik AutoML – you are prompted to select data from the Qlik Catalog. Though you will see data that has been registered in the catalog, it is important to note that not just any data can be used with Qlik AutoML. Historical data ( data you normally use for analytics and reporting with dimensions, measures, time periods, etc. ) is generally not suitable for machine learning. Why? Data you’re accustomed to using in analytics apps is captured all at the same point in time (current, end of day, end of month, etc.). With machine learning and predictive analytics, the dataset needs to contain an outcome we care about and a set of attributes we call features (typically your dimension fields / values)  that existed prior to that outcome. That way, a machine learning model can learn the patterns that led to that outcome and therefore identify them in advance when we apply the model to live or current data. That being said – you must select data from the catalog that has been prepared with the main ingredients needed to train machine learning models. 

Stay tuned for Tip #3 where I will briefly cover architecting the data set and the needed ingredients.

Qlik AutoML Helpful Resources:

• Sample Data Page with basic creation of a Qlk Sense predictive application - walkthrough 
• Qlik AutoML Playlist (on YouTube)
• Qlik Help Documents - (will go live 8/16 this week)
• Qlik Continuous Classroom Learning Modules
• Qlik AutoML in the Qlik Community