Power BI AI Features and Key Influencers Visual

On 21 Sept., 2019

Automate AI with Power BI Microsoft has released AI features for PowerBI and here are some of them.

Power BI AI Features and Key Influencers Visual

Automate AI with Power BI

Microsoft has released AI features for PowerBI and here are some of them.

And no additional tools to be purchased: Just your good old Power BI with Premium Capacity.

What is in it?

  • Cognitive Services to tap into existing, powerful services of image recognition, sentiment analysis, and much more.
  • Auto-ML to train a novel machine learning model for your data. AutoML automatically searches and tunes the best algorithm for your dataset.
  • Azure ML Integration to integrate the machine learning models you have developed in Azure Machine Learning Studio, directly to your Power BI workspace.

Power BI AutoML in action

Let’s say that you have an online retail store, and you want to predict if your customers are going to purchase your products based on their interactions. Such a prediction is very important, as you can:

  • Dynamically adjust your website’s response to the customers – especially convince the customers that may otherwise change their mind the last moment.
  • Improve your marketing campaigns to bring the best to them.

This is clearly a machine learning task, because there are so many features you will need to take into account in order to have an accurate understanding of the purchase behaviour. Here, Power BI comes in: plug in your data, and Power BI’s dataflow will find the best algorithm that predicts the behaviour.

Power BI’s Machine learning capabilities comes in various flavours: from classification, i.e. finding the right categories for your data (e.g. successful purchase or not) to regression and forecasting (e.g.: based on our existing performance, what will be our sales numbers in a couple of months?). For this use case, we will go with classification.

Your data is split into train and test partitions – train is to let your model discover patterns, and test is to validate if the discovered patterns indeed reflect the reality your customers’ behaviours.

After the model’s training is done, Power BI shows you a report that unveils the evaluation process of the algorithms running in the background. No more black magic – only clear interpretable results.

Furthermore, we are also able to see the key influencers for the purchases. It seems like:

  • Page rank of your website/product’s page (pagevalues),
  • The month when the potential customers visit (Nov)

Are the top influencers of the purchases. But wait, there is more!

The operating system of your customer (Operatingsystems(1)) have a very high true outcome / false outcome ratio. It means that the operating system of the customer’s device (OS X? Windows? Android?) have a very predictable impact on the purchase behaviour.

Just like that, you now have some actionable insights to improve your sales: render your page more accessible to search engines, concentrate your marketing on the month of November, and include elements that draws inspirations from the usage of operating systems!

Great, but what if I don’t want to use Premium Capacity?

You still have an alternative: you can provision Power BI Embedded in your Azure cloud tenant, and associate this with the Power BI workspace in which you want to unleash the AutoML capabilities. One particular advantage of this approach is that you can pause the execution of your PBI Embedded resource, saving further costs for your organization.

Stay tuned with us for the updates, and even better, contact us so that we can discuss how can Power BI’s AutoML features can address your business needs.

What else can we tell about Power BI’s AI capabilities?

Besides this, the “Key Influencers” feature is also available in Power BI Desktop. The key influencer visual is a great choice:

  • To see which factors impact the metric being analysed.
  • To contrast the relative importance of these factors.

We’ve applied this feature on our recent supply chain analysis report, and identified the key influencers that cause production levels lower than expected. Raw material costs indeed influence the production levels, but not as much as the overload rate of the production machines and the companies that supply the parts of the machines. Not a good day for the vendor company A, whose machine components are the biggest influencers for low level of productions.

Would you like to learn more? Contact us and we’ll bring your reports to life with Power BI and its AI features!

If you are interested in PowerBI Consulting  AI related solutions then, please contact Prometix – enquires@prometix.com.au