Using the Model Builder and AutoML in Microsoft ML.NET

Step-by-step guide for creating, training, evaluating and consuming machine learning models powered by ML.NET

Transcript

1. Using the Model Builder and AutoML in Microsoft ML.NET
2. >whoami Senior .NET Software Dev ML Researcher Technical Trainer Macedonian .NET Community member https://www.linkedin.com/in/miodragcekikj/ https://medium.com/@cekikjmiodrag https://www.researchgate.net/profile/Miodrag-Cekikj https://speakerdeck.com/mcekic https://github.com/mcekikj
3. “ Life is not about getting and having. Life is about giving and being. Hajan Selmani
4. Session Roadmap 1 3 5 6 4 2 What is Machine Learning? Model Builder and AutoML ML.NET in action (Demo time) ML.NET - 4Ws Journey Caveats Key Takeaways
5. 1. What is Machine Learning? Let’s get rolling folks!
6. Machine Learning Represents an application of the Artificial Intelligence (AI) focused on the use of data and algorithms allowing the software applications to learn and improve without being explicitly programmed.
7. ML versus Classical Programming https://tinyurl.com/y6dpr3ww - Sravya Reddysetty, Traditional Programming vs Machine Learning Related resources: https://tinyurl.com/yk6br74u https://tinyurl.com/35etdrdh
8. https://tinyurl.com/ym8wwu2b - Algorithmia 2021 enterprise trends in machine learning, published in Forbes AI/ML trends
9. https://tinyurl.com/ym8wwu2b - Algorithmia 2021 enterprise trends in machine learning, published in Forbes AI/ML trends
10. 2. ML.NET – 4Ws What, Where, When, Why…
11. ML.NET An open source and cross-platform machine learning framework that democratizes the Machine Learning for .NET developers
12. From zero to hero… ◎ Research project for text mining and search ◎ Learning code - Microsoft internal framework ◎ ML.NET – 7 May, 2018 as external rebranded version ◎ Part of the .NET release schedule (1.7 just released) ◎ GitHub repository: https://github.com/dotnet/machinelearning/ ◎ Roadmap: aka.ms/mlnet-roadmap
13. ML.NET is actually .NET ☺ ◎ Empower and stay in the .NET ecosystem ◎ C# or F# coding skills ◎ Visual Studio or VS Code preferences ◎ No deep data science or ML related experience needed ◎ Experiment to production concept
14. 14 Industry Challenge Technological gap between the data scientists and software engineers Photo by Jerome Dominici from Pexels
15. Deployment horizon Console Mobile Web Desktop .NET framework (version 4.6 or above) & .NET Core (version 2.0 and above)
16. 3. Model Builder and AutoML Prototype & Automate
17. Model Builder ◎ Wizard built on top of the ML.NET SDK ◎ Best trainer for the data and specific scenario ◎ Can be different as the data changes over time ◎ Allows quick prototyping ◎ Visual Studio, VS Code and CLI support Microsoft documentation reference: https://tinyurl.com/4dr6csw7
18. AutoML (Automated Machine Learning) ◎ Automates the model experimentation process ◎ Not specific just for the ML.NET and .NET ecosystem ◎ Raw dataset to deployable ML model pipeline ◎ Model Builder is using it behind the scenes (via UI/CLI) Microsoft documentation reference: https://tinyurl.com/mwx5s6fc
19. 4. Journey Caveats Widen you gaze
20. 20 Photo by Steve Johnson from Pexels
21. First things first… ◎ ML would not magically solve our problem ◎ Understand the business domain ◎ In-depth data analysis ◎ Guide ML toward right direction ◎ Begin with KISS principle (keep it simple, stupid)
22. 22 Photo by Pixabay from Pexels
23. 5. ML.NET in Action Demo time
24. Domain and scenario introduction ◎ Lead Scoring/Lead Decision ◎ Shared sales and marketing methodology used to rank leads to describe their potential of sales readiness to the company ◎ Technique of assigning different values to the company`s leads database guiding the marketing and sales teams through conversion to “hot leads” and official customers/clients
25. Dataset ◎ Lead Scoring Dataset ◦ Kaggle Source: https://www.kaggle.com/amritachatterjee09/lead-scoring-dataset ◦ Prerequisite: Data processing (Data Cleaning, Exploratory Data Analysis, Data Preparation) ◦ Regression & Classification (Supervised ML) approach in ML.NET
26. Supervised ML Supervised Learning by Jorge Leonel on Medium: https://tinyurl.com/23su62xx
27. Why ML in the story? ◎ Digital marketing, lead generation, and sales teams generate a lot of lead data ◎ Stored in a predetermined structured format on single/multiple platforms ◎ Use the complete history of generated data (labelled datasets as a prerequisite for the supervised learning approach)
28. Prerequisites ◎ At least Visual Studio 2019 16.10.4 or later/.NET Core 3.1 SDK or later ◎ ML.NET Extension - installing and enabling the Model builder ◎ ML.NET Model Builder UI extension tool ◎ Let`s start… ☺
29. 6. Key Takeaways Keep the ball rolling!
30. Ready, Steady, Go ◎ Effective prototyping and experimentation ◎ Embracing the Automated ML approach ◎ Open source & cross-platform ◎ Flexible and extensible ◎ Trusted and proven at scale
31. Thanks people! Let`s discuss, let`s connect!
32. Session related resources ◎ Towards Data Science via Medium article - https://tinyurl.com/3d9nxx24 ◎ GitHub repository - https://github.com/mcekikj/yes2021-ml-net ◎ Presentation via Speaker Deck - https://tinyurl.com/mr45msrb
33. Extra Resources ◎ Kaggle - https://www.kaggle.com/datasets ◎ The Boston Housing dataset - https://tinyurl.com/yckwj69v ◎ Iris dataset - https://tinyurl.com/y5w5evsz ◎ Titanic dataset - https://tinyurl.com/4b7ncvz8
34. Credits More info on how to use this template at www.slidescarnival.com/help-use-presentation- template This template is free to use under Creative Commons Attribution license. You can keep the Credits slide or mention SlidesCarnival and other resources used in a slide footer. More info on how to preview and download other images at https://www.pexels.com/ The images are free for use according the Legal Simplicity vision of Pexels.