Azure Machine Learning Challenge_Speakers Presentation.pptx

Machine Learning with Azure

1. Date: December 20, 2019 At: Computer Lab, Central University of Punjab, Bathinda Ease of Applied Machine Learning using Microsoft Azure Learn about the ease of using Azure in Machine Learning and participate in ML challenge. Also create the web service for applying machine learning to real life.

2. Have you ever wanted to predict the future?

3. What is Machine Learning? Branch of computer science in which a computer "learns" from data in order to perform predictive analytics Credit-card fraud detection Online shopping recommendations Self-driving cars and more Supervised learning Regression and classification Unsupervised learning Clustering

4. Machine Learning uses historical data to make predictions

5. Data Science = Data  Insight  Experience  Action

6. Machine Learning in Action

7. Machine Learning in Action

8. Machine Learning in Action

9. Microsoft and Machine Learning Modified from http:// pulsweb.fr /predict-wine-quality- azureml

10. I spent last semester building a Neural Network model in Python, and I just did the same thing in 10 minutes with Azure ML

11. Using Azure Machine Learning Studio, you can explore the depth of Machine Learning without a degree in statistical analysis

12. Machine Learning studio Visual editor for composing, testing, refining, and deploying machine-learning models Includes hundreds of modules Includes common algorithms for classification, regression, and more Supports numerous input formats Supports R and Python Machine learning for the masses

13. The Machine Learning Process From "Introduction to Microsoft Azure" by David Chappell

14. Machine Learning uses historical data to make predictions Using Azure Machine Learning Studio, you can explore Machine Learning without coding or a degree in math… …but you can code with it.

15. Challenges?

16. Start with the data "Clean" the data and train the model Create the web service

17. Where do you get the data? Use any of the 39 sample data sets built into Azure ML Studio Create new data sets by uploading CSV or TSV files Use the Import Data module to import data from Web URLs, SQL databases, Hive queries, OData feeds, blob storage, and other sources. We here use the data: http://tiny.cc/workshopdata

18. The better insights, the better results!

19. Demo! Launch Azure ML Studio Load a dataset Visualize the data

20. Start with the data "Clean" the data and train the model Create the web service

21. Data almost always requires preparation or "cleaning" Azure ML Studio provides several modules to help with this task, including Select Columns from Dataset , Clean Missing Data , and Remove Duplicate Rows.

22. Training the model involves using a learning algorithm to analyze the data and find patterns ML Studio includes 25 modules implementing common ML algorithms: Linear Regression , Multiclass Neural Network , One-Class Support Vector Machine , and many more.

23. Simple (Univariate) Linear Regression Regression line represented by an equation of the form Y = b + b 1 X where Y is the dependent variable Error between actual and computed output minimized using least-squares or gradient-descent method

24. http:// aka.ms / MLCheatSheet

25. Which algorithm to use depends on the type of prediction you want to make Algorithm type Type of prediction Classification Predict a category, e.g. what income range you fall into, which hockey team you cheer, or which political party you prefer Regression Predict a value, e.g. someone’s income or the price of gas Anomaly detection Find anomalies in the data, e.g. credit -card fraud detection

26. After you train the model, check the accuracy Use Score Model to "score" the model, and Evaluate Model to visualize the results

27. Visualize the output of Evaluate Model to assess results You may need to make changes to improve the results

28. Demo! Prepare the data Train the model Check the accuracy

29. How can you improve accuracy? More rows of data. More or fewer columns of data (choose the ones that are most statistically significant and eliminate ones that aren’t). Cleaner data (remove or replace missing or incorrect values, minimize imbalance, normalize non-normalized values, remove outliers, etc.). Different algorithms. Different algorithm parameters.

30. ML Challenge restrictions #1 No deleting rows of data (or fixed values). Why? You’re changing reality :) #2 No using data that can’t be inputted by a user. Why? Users book a flight (date + destination)

31. Hint #1 garbage in, garbage out

32. Hint #2 minimize imbalance Hint #2 Minimize imbalance

33. Hint #3 Sometimes you must go backwards to go forward

34. Hint #4 try more algorithms and parameters Hint #4 Try more algorithms and parameters

35. Hint #5 more rows of (clean) data = better results Hint #5 More clean data

36. Start with the data "Clean" the data and train the model Create the web service

37. Once you have a trained model, you can deploy it as a web service Create a predictive model Deploy it as a web service Call it from your code to ask for predictions

38. Demo! Create a predictive model Deploy the web service Test the web service

39. You can also use someone else’s trained models to do cool stuff!

40. Use the Computer Vision API to analyze images for adult content, generate captions, and more

41. Use the Text Analytics API to identify sentiment expressed in text (e.g., Twitter feeds) “Thanks to Text Analytics…we are able to incorporate guest sentiment into our actionable guest feedback platform that delivers a comprehensive view of guest satisfaction and server performance.” — Al Pappa , Head of Business Intelligence, Ziosk Language Detection Topic Detection Key Phrase Extraction Sentiment Analysis

42. ! Time to clean now! Delete the resources which might hurt in background. Keep an eye on the computing resources.

43. And now finally ..

44. Thank you! 

45. Thank you