SparkBeyond Discovery Platform V1.33

  • profile-pic_13March19-e1580118780255 (1)

We’re thrilled to announce the newest version of the SparkBeyond Discovery platform, 1.33. We’ve made some improvements to the feature search process and overall user experience, giving you the ability to do even more with your data.

Here’s what to expect:

Streamlined integration of internal and external data 

We’ve made it easier for you to connect your internal data with external sources. With this new release, you can now pinpoint the most efficient methods of connecting your data based on your target definition. As a result, you will be able to uncover novel ways to get more out of your data as a whole.

The system will highlight relevant external datasets based on your target and data types

Control the final feature selection process

You know your data better than anyone else. That’s why we’re giving you better control over the feature search process and the results of automated feature engineering.

We’ve made it simpler for you to decide which features to keep and which to drop from your feature set. Excluded features will be completely removed from your search space, so they won’t be seen in future iterations.

In the next release, you’ll be able to exclude further elements like functions from a selected column or from the entire pipeline.

Build and refine models better

Take advantage of a newly designated space to build and refine your models. We’ve also added a brand new model evaluations tab. This tab provides essential information about your model performance including Curves plot, Confusion Matrix, Evaluation metrics bar, Threshold Slider, and more.

With these new tools, you can now compare Train and Test results side-by-side, share professional plots with business stakeholders, and find and analyze optimal threshold values.

For more information, click here

A newly designated space to build and refine your models


Uncover meaningful insights

We have implemented new functionalities to allow you to automatically identify different types of phenomena in your data. 

Time Intervals Context

It’s now easier to find insightful hypotheses about objects with a well-defined lifetime in your analysis. For example, if your data includes promotions or insurance policies containing start and end dates, the system is now able to express various hypotheses such as: The number of ongoing promotions; The time until the first ongoing promotion ends; and The attributes of ongoing promotions.

Improved Time Series Forecasting with Prophet

With our new time series forecasting capabilities based on the Prophet algorithm, you can now add predictive features based on time series components (e.g. trend, seasonalities, and events). You’ll also be able to predict their specific effects on the target, such as the time of year on a specific store’s sales. 


To learn more about 1.33 functionalities, click here.

Ready to upgrade? Contact your system admin.

If you have any feedback or comments please reach out to

Related Posts

Scaling-up analytics: what major CPG & e-commerce are doing right
The Game Changers:
AI in Finance
The Game Changers: AI in Banking

Intelligence in your inbox

Subscribe to our blog for SparkBeyond’s latest news and insights.