Upload
vanngoc
View
245
Download
0
Embed Size (px)
Citation preview
Microsoft Ignite 2015
2015 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
9/3/2015 5:27 PM
1
A Journey into Azure Machine Learning with R
Leila Etaati
M234
2
Who Am I ?
Leila Etaati
Data Mining and BI Consultant
Speakers in Many Microsoft SQL Server Conferences (SQL Rally, Code camp and SQL Saturday )
10 Years experiences in SQL Server
Co-Lead Auckland BI User Group
PhD in Information System Department, Business School University of Auckland
Lecturer and Tutor of BI and database
@leila_etaati
Microsoft Ignite 2015
2015 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
9/3/2015 5:27 PM
3
Agenda
4
Introduction to Machine Learning
What is Azure ML
Azure ML Demos
Azure ML with R
Facts about Azure ML
Introduction to Machine Learning
5
I believe over the next decade computing will become even more ubiquitous and intelligence will become ambient this will be made possible by an ever growing network of connected devices, incredible computing capacity from the cloud, insights from big data, and intelligence from machine learning.
Microsoft Ignite 2015
2015 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
9/3/2015 5:27 PM
6
If you invent a breakthrough in Artificial Intelligence, so machines can learn, that is worth 10 Microsoft
Microsoft Ignite 2015
2015 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
9/3/2015 5:27 PM
7
What is Machine Learning ?
The goal of machine learning is to build
computer systems that can
adapt and learn from their experience.
-Tom Dietterich
Microsoft Ignite 2015
2015 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
9/3/2015 5:27 PM
8
Information Optimization
Value
Difficulty
What
Happened ?
Descriptive
Analytics
Diagnostic
Analytics
Predictive
Analytics
Prescriptive
Analytics
Why did it
Happen?
What Will
Happen?
How can we
Make it Happen?
9
Mortgage
Applications
Pattern
Recognition
Health
Insurance
Fraud
Detection
Airline
Flights
Web Search
Page result
Example of Using Machine Learning
10
Example of Using Machine Learning
11
Microsoft Speech Recognition Control
Microsoft Search Engine
Microsoft Xbox and Kinect
12
Machine Learning
13
Should I used Machine Learning
Predication is Small
part of experiences
No Past data
Many Rules govern
Experience
Automated Predication is Core
Lots of History
Magic numbers in current prediction system
14
Machine Learning Concepts
Data
Model
Parameters
Learning
Prediction
15
Steps
First-Understand the Business Domain
Second-Understand the Business Problem
Third- What is the Right Data, Right
Column and Right Algorithm
Last-Combine Knowledge With
Machine Learning
16
Steps to Build a Machin Learning
De/Refine
Business Rule
Sales More
Find Fraud
Find Potential Customers
Collect and Clean
Data
ETL Process
Split Data
Train Model
Test Model
Performance
Score Model
What type of Data
% for
Training
Choose Model
% for
Testing
17
Evaluate
Models
De/Refine
Business Rule
Collect and Clean
Data
Split Data
Choose the Best Model
Model A
Model B
18
What is Azure ML
19
Data
scientist
20
Microsoft Azure Machine Learning
Reduced Complexity
Access Through Web Browser, no need to install any thing
Collaborate work with anyone
Visual composition, easy to use, No Coding
Good storage of Algorithm (Use in Bing search, Xbox..)
Have good support for R studio, Python and Jupyter notebook
Load Data From Different Location
Clean data
Machine
Learning
Algorithms
R and
Python Language
Web services
21
Cortana Analytics Suite successful experience with Azure MLhttps://www.youtube.com/watch?v=YxmAEMmwXYU
22
Azure ML Demoes
23
Demo: Azure ML Environment
Leila Etaati
Microsoft Ignite 2015
2015 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
9/3/2015 5:27 PM
24
Demo:Classification Problem Example
Leila Etaati
25
Titanic Classification Algorithm
Survived or Not
Titanic sank on15April 1912 after colliding with an iceberg
26
Azure ML using R code
Leila Etaati
Microsoft Ignite 2015
2015 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
9/3/2015 5:27 PM
27
Demo: Azure ML with R
Leila Etaati
28
Fact about Azure ML
29
Azure ML Algorithm
30
Regression Algorithm
Classification
Algorithm
Discrete
Variables
Customer Preferences
Martial Status
Married, Divorce, Single
Income
More than 50K or
Less Than 50K
Continues
Variables
Income, Sales, Profit
Match Data Type
31
Clustering Algorithm
Regression Algorithm
Classification
Algorithm
Descriptive
Analysis
Predictive
Analysis
Predictive
Analysis
Descriptive and Predictive Analysis
1-Which other customers have similar preferences to this one?
2-What are the most common patterns in gasoline price changes?
1-When will this customer make another purchase?
2-How many new followers will I get next week?
1-Will this customer click on the top link?
2-Which offer should this customer receive?
32
8.19999999999999933.21.41.2Sales1st Qtr2nd Qtr3rd Qtr4th Qtr
Clustering Algorithm
Regression Algorithm
Classification
Algorithm
Supervised
Learning
33
34
SSAS vs Azure ML
35
SSAS Vs Azure ML
FeaturesUsabilityCostSupportEnd to end ProductCanned algorithmNot possible to change algorithmDMX CodeMore Visual Excel VersionIts not easy to startAll users can useIf you purchase SQL Sever: FreeFew books and small online community
Current and up to date algorithmIntegration with R and PythonCloud baseREST formatHard to interpret Drag and Drop UICustomize the AlgorithmFree version, limited optionsMore online Community
36
Azure ML Pricing
37
Questions
38
Related Ignite NZ Sessions
Using Azure Machine Learning to predict Trade Me auction prices Ballroom 1 (Crowne Plaza)Wed 1:45pm
Windows 10 + Azure AD + Intune = full desktop management and provisioning in the cloud New Zealand 1 (SKYCITY)Fri 9:00am
Data Patterns for the Cloud Ballroom 1 (Crowne Plaza) Wed 10:40am
Azure Machine Learning: From Design to Integration Marlborough (SKYCITY) Fri 10:40am
Find me later at
Hub Happy Hour Wed 5:30-6:30pm
Hub Happy Hour Thu 5:30-6:30pm
Closing drinks Fri 3:00-4:30pm
1
2
3
4
39
Resources
TechNet & MSDN Flash
Subscribe to our fortnightly newsletter
http://aka.ms/technetnz http://aka.ms/msdnnz
http://aka.ms/ch9nz
Microsoft Virtual Academy
Free Online Learning
http://aka.ms/mva
Sessions on Demand
2014 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
9/3/2015
40
Complete your
session evaluation
now and win!
41
9/3/2015 5:27 PM
2013 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
2015 Microsoft Corporation. All rights reserved.
Microsoft, Windows and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
9/3/2015 5:27 PM
42
2014 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.