Synapse Data Science – in MS Fabric

Synapse Data Science

Synapse Data Science is in many ways the newest kid on the Fabric block.  Most of the buzz and headlines in the World of Data are concentrated here. This is where you start to build your own Machine Learning models, and first benefit from your own organizational AI.

Data Science and the 4 classes of Business Analytics

Data Science opens up the possibility of “Predictive Analytics”, one of the 4 classes of Business Analytics

  • Descriptive – What happened – traditional reporting
  • Diagnostic – Understanding – Why did it happen?
  • Predictive – Using models – understand what might happen?
  • Proscriptive – Compare possible outcomes – what should happen?

4 types of Business Analytics – Harvard Business School

Questions Data Science can answer

It can help you by answering questions like:

  • Which customers will churn in the coming months?
  • Which of these transactions is fraudulent?

Preparing to benefit from Data Science

As with other uses of your data, Synapse Data Science relies very strongly on having well-managed and reliable data to work on.  The “GIGO” (garbage in, garbage out) principle applies when source data isn’t of sufficient quality.

Data Science allows you to run machine learning models against your data using popular languages like Python and “R”, and then compare the outputs of different runs so you can understand which results are most useful to the business.

Before you start implementing data science, most experts recommend that you prepare “Decision Grade” data. This data has been cleaned, verified and automated so when used in a learning model – you learn the right things.

Easy-start in Synapse Data Science

And as ever in Fabric there’s some easy-start options, where you can take popular ML models and run them against your data without needing to write any code.

Cost-control in Fabric

From a cost-control perspective its worth remembering that good machine learning often requires a large amount of data: “big data”, and that both the volume of data required and the amount of processing power can quickly overwhelm a regular “shared” capacity.  So while you can trial ML models on a shared capacity, this is the area most likely to push you towards Microsoft Fabric’s top-tier pricing model, their “Premium Capacity” where you pay for a number of processing cores or “Capacity Units” but can run any number of users on them.

There’s more on how this works on Comset’s “Fabric Pricing” page.

Overview of Data Science in MS Fabric

Data Science opens up the possibility of Predictive Analytics

Ask Comset

Our team are ready and waiting to help you navigate the world of Data. Please get in touch.