Getting started with Data Science in Microsoft Fabric
Looking to get started with Data Science in Microsoft Fabric? In this session, we'll walk through how to get started by creating a Jupyter notebook, download sample data from our lab, create a ML forecast with Prophet, and cross-validate and evaluate the model. We'll then store the model in MLflow and see how to use it both in the notebook and in the Fabric workspace.
The notebook is available for download in the data science module in our lab at https://aka.ms/fabricrealtimelab -- this module is designed to be stand-alone, so it can be completed without finishing other modules in the lab.
Links from video:
Real-time Lab with Notebook https://aka.ms/fabricrealtimelab
Prophet https://facebook.github.io/prophet/
Chapters:
00:00:00 Introduction
00:00:46 Eventstream ingestion to Lakehouse
00:01:24 Overview of the Jupyter Notebook
00:03:18 Download historical data
00:03:55 Load CSV into dataframe and process
00:07:35 Quick look at Data Wangler
00:09:14 Training the model (Prophet)
00:16:15 Cross-validation
00:24:12 Log the model with MLflow
00:25:28 Loading the model and generating predictions
00:29:52 Comparing the model vs actual
00:31:27 Scikit-learn metrics
00:32:29 Additional MLflow
00:33:02 Looking at the models in the UI