Microsoft Fabric Data Science project tutorial

Title image
6th Jan 2024

Microsoft Fabric functionalities allow you to leverage data science capabilities while developing machine learning models for being applied on your data to predict future outcomes, like for example when trying to do forecasting for Bitcoin prices, as I show you how to do that in this tutorial.

This data science project leverages a gold layer data based on historical Bitcoin prices placed in a fact table within a star schema data model on a fabric data warehouse developed in the first part of an end-to-end Fabric project, that on top of that I built a semantic model for visualizing trends on Bitcoin prices. So this time I wanted to predict future Bitcoin prices as well as visualize them against actual values in the Power BI report.

The workflow I followed in this tutorial is based on the Microsoft learn documentation, so it involves the following parts...

  1. Retrieving some data from a semantic model by using the semantic-link library in a fabric notebook
  2. Running some machine learning models in it by using Python, where I do some EDA (exploratory data analysis) of Bitcoin prices to learn and get some findings on prices seasonality.
  3. MLflow to monitor the different forecasting models developed in a fabric experiment to be able to select the best-performing model. 
  4. Store the predicted values in a delta table in the lakehouse. In my case as well I placed in the gold layer as a new table within the star schema in the data warehouse so we can add this new table to the existing semantic model.
  5. Visualize the predicted values against the actuals in the Power BI report.

Please find the whole explanation to follow along with the project tutorial in the following video. Hope you find it interesting. Cheers, Adrian

About me
Adrian Rodriguez

Spaniard working in the data field since 2013 in different industries as BI developer, Data Analyst as well as Data Scientist. I consider myself a data analytics passionate.