Azure machine learning 101 - Part 1
Download as PPTX, PDF1 like220 views
The document outlines a plan for improving performance analytics through techniques such as data wrangling, retrieval, storage, data mining, and machine learning. It includes steps for defining, training, validating, and deploying machine learning models, as well as discussing concepts like classification, decision trees, entropy, and feature importance. Additionally, it refers to resources for understanding machine learning algorithms and classifier design.
Azure machine learning 101 - Part 1
- 8. Data wrangling (munging), retrieval + storage Data mining & machine learning Statistics Big data Data science
- 11. Source: Data Science For Beginners - 5 Questions Data Science Answers by Brandon Rohrer
- 12. Is this A or B?
- 13. Is this Weid?
- 16. What should I do now?
- 20. 1. Define 2. Train 3. Validate scoring evaluating Explore Deploy …visualise and study …deploy as a (web) service 5. Update and revalidate
- 23. 23
- 25. 95 230
- 26. Gini 𝐺𝑖𝑛𝑖 = 1 − 𝑗 𝑝𝑗 2 Where p is the probability of having a given class in your dataset Entropy 𝐸𝑛𝑡𝑟𝑜𝑝𝑦 = 𝑗 −𝑝𝑗 ∗ 𝑙𝑜𝑔2 𝑝𝑗 Where p is the probability of having a given class in your dataset percentage percentage
- 27. 100 Points Split on Elevation 4060 3010 Split on PriceSplit on PSF Split on PSF 2 3 4 1 Split # of data points Feature Information gain (IG) Data points * IG 1 100 Elevation 0.3 30.00 2 60 PSF 0.4 24.00 3 40 Price 0.2 8.00 4 30 PSF 0.1 3.00
- 28. 100 Points Split on Elevation 4060 3010 Split on PriceSplit on PSF Split on PSF 2 3 4 1 Feature Importance Normalized Elevation 30.00 0.46 PSF 27.00 0.42 Price 8.00 0.12
- 30. Lior Rokach
- 33. select the simplest one keep all theories
- 35. Final Class: +1
- 36. B AGG ING
- 37. T m n
- 38. Backups
- 44. 3. Modify the training set some combination
Editor's Notes
- #18: I want to start by grounding this discussion in Microsoft’s mission because everything we’re going to talk about today begins and ends with our mutual goal of empowering our customers and our communities. To that end we have three bold ambitions, and they form the foundation of the evolution of the MVP Award. Building the intelligent cloud. We want to have the richness of the cloud infrastructure that accommodates for the complexities and the diversity of the way the world works today. Our goal is to build the most comprehensive cloud infrastructure and support hybrid computing in its purest form so that it provides flexibility. And then build out a rich data platform on top of it for building new systems of intelligence that drive business transformation. Create more personal computing. This is where we focus on creating a more natural interface and building experiences that are about the mobility of the experience, not the mobility of the device. Increasingly the issue of our time is going to be privacy and being in control of what data you share, how that data is being used. Windows 10 marks a huge milestone on this journey of more personal computing where it will span with one consistent experience for users, developers, and IT administrators from across Internet of Things to these holographic computers. Reinvent productivity and business process. The intersection between productivity and communication and collaboration and business process. This is where Office 365 and Dynamics in particular come together to enable that new way of working, the new workflows of the enterprise.