Predicting the future in 10 minutes or less - an introduction to Azure Machine Learning

Transcript

1. Predicting the Future
   in 10 Minutes
   or Less
   

   View Slide

2. ...a 45-minute presentation by
   @vladiliescu
   

   View Slide

3. What do you mean, “Predict the future”?
   

   View Slide

4. Predict the future
   

   View Slide

5. Predict the future grades of a student
   

   View Slide

6. Predict the future grades of a student
   at math
   

   View Slide

7. Predict the future grades of a student
   at math using...
   

   View Slide

8. a) Divination
   

   View Slide

9. b) Statistics
   

   View Slide

10. c) Machine Learning
    

    View Slide

11. c) Machine Learning
    A way to write programs that we don’t exactly
    know how to write.
    

    View Slide

12. c) Machine Learning
    A way to write programs that we don’t exactly
    know how to write.
    

    View Slide

13. Predict the future grades of a student
    at math using machine learning…
    

    View Slide

14. Apache Spark MLlib
    Microsoft Azure ML Studio
    Google Prediction API
    Google TensorFlow
    Amazon Machine Learning
    Samsung Veles
    

    View Slide

15. Microsoft Azure ML Studio
    

    View Slide

16. What have we picked so far?
    ✔ A well-defined problem
    ✔ An approach for predicting the future
    ✔ A machine learning framework
    

    View Slide

17. What are we missing?
    

    View Slide

18. The Data
    

    View Slide

19. UCI Student Performance Data Set
    

    View Slide

20. https://archive.ics.uci.edu/ml/datasets/Student+Performance
    

    View Slide

21. G3 - final grade (numeric: from 0 to 20, output target)
    

    View Slide

22. Ready?
    

    View Slide

23. Steady?
    

    View Slide

24. GO!
    

    View Slide

25. Open the Azure experiment here
    https://gallery.cortanaintelligence.com/Experiment/Predict-Student-Grades-at-Math
    (or take a look at the following slides)
    

    View Slide

26. View Slide

27. View Slide

28. View Slide

29. View Slide

30. View Slide

31. View Slide

32. View Slide

33. View Slide

34. View Slide

35. View Slide

36. View Slide

37. View Slide

38. What have we done so far? (continued)
    ✔ Trained&deployed the simplest model we could
    ✔ Consumed the web service using Excel
    ✔ Learned how to evaluate a trained model
    

    View Slide

39. Thanks for watching!
    Vlad Iliescu
    ↪vladiliescu.ro
    

    View Slide