This session is about learning how to use Microsoft Azure Machine Learning with the devices in order to detect data patterns. This session will cover an introduction to Machine Learning, and different algorithms used to detect data patterns. The algorithms discussed will be nearest neighbor, probabilistic learning, decision trees, and neural networks. It will also cover data that comes from devices like the Kinect for Windows device. The session will show basic demos and data coming from the device. The session will then drill down into how to incorporate Azure Machine Learning features into an application to detect data patterns in real time.
Using Azure Machine Learning to Detect Patterns in Data from Devices
1. Sponsored & Brought to you by
Using Azure Machine Learning to Detect
Patterns in Data from Devices
Dwight Goins
https://www.linkedin.com/pub/dwight-goins/1/836/591
3. AGENDA
• 5 Reasons to use Azure ML with Devices
• IoT and Cloud based Devices
• Need for processing data
• Machine Learning (ML) Overview
• ML Approaches
• Implementing ML Algorithms
• Microsoft Azure Machine Learning
• Using a Device with Azure ML
4. 5 REASONS
• Recognize Data Patterns
• Predict Actions and Events
• Find similar data - group
• Rank data
• Quickly Experiment with Innovative Ideas
17. USING THE KINECT DEVICE WITH MAML
• Kinect Device has 4 sensors
• Color
• IR
• Depth
• Microphone
• Generates A LOT of data
• Demo
18. STEPS TO DETECT DATA PATTERNS
• Record Tests into Spreadsheets
• Use Azure ML
• Expose Azure ML Web Services
• Build Application
19. APPLICATION PROCESS FLOW
Device
• IOT Devices – Events &
Data
• Kinect – Events & Data
Connectivity
• Event Hubs
• Service Bus
REST/WCF
Storage
• SQL DB
• BLOB
• Doc DB
Analytics
• Azure ML
• Steram Analytics
• HDInsight
• Data Warehouse
Presentation/Action
• App Service
• PowerBI
• Notification hubs
• Mobile Services
• BizTalk services
20. RECAP 5 REASONS
• Recognize Data Patterns
• Predict Actions and Events
• Find similar data - group
• Rank data
• Quickly Experiment with Innovative Ideas
21. CALL TO ACTION
• Studio.azureml.net – Register and start playing
• Windows 10 – IoT devices
• KinectForWindows.com
• Dgoins.wordpress.com
• Nimbo.com
22. REFERENCES
• R – http://r-project.org
• ENCOG – HeatonResearch.com/encog
• Azure ML – http://Studio.azureml.net
• MVP Virtual Conference - aka.ms/mvpvconf
• Dr. James McCaffrey – MSDN Articles https://msdn.microsoft.com/en-
us/magazine/hh975375.aspx
• Dwight Goins Blog – http://dgoins.wordpress.com
Using Azure Machine Learning to Detect Patterns in Data from Devices
ML recognizes data patterns – Devices have a lot of data you need some engine to identify and classifly.
ML can help predict actions or guess estimations (regression) based on data/events – Devices gather action data and events
ML can find hidden features and help classify data – Sensors has a lot of data
IoT is a buzzword for devices with many sensors – ML can make sense of the data, group and rank data.
Use ML as a testbed for POC’s and experimental projects and solutions and help reduce or logically deduct conclusions
(Classification, Regression, Ranking, Clustering, Dimensionality)
ML recognizes data patterns – Devices have a lot of data you need some engine to identify and classifly.
ML can help predict actions or guess estimations (regression) based on data/events – Devices gather action data and events
ML can find hidden features and help classify data – Sensors has a lot of data
IoT is a buzzword for devices with many sensors – ML can make sense of the data, group and rank data.
Use ML as a testbed for POC’s and experimental projects and solutions and help reduce or logically deduct conclusions
(Classification, Regression, Ranking, Clustering, Dimensionality)
ML recognizes data patterns – Devices have a lot of data you need some engine to identify and classifly.
ML can help predict actions or guess estimations (regression) based on data/events – Devices gather action data and events
ML can find hidden features and help classify data – Sensors has a lot of data
IoT is a buzzword for devices with many sensors – ML can make sense of the data, group and rank data.
Use ML as a testbed for POC’s and experimental projects and solutions and help reduce or logically deduct conclusions
(Classification, Regression, Ranking, Clustering, Dimensionality)
ML recognizes data patterns – Devices have a lot of data you need some engine to identify and classifly.
ML can help predict actions or guess estimations (regression) based on data/events – Devices gather action data and events
ML can find hidden features and help classify data – Sensors has a lot of data
IoT is a buzzword for devices with many sensors – ML can make sense of the data, group and rank data.
Use ML as a testbed for POC’s and experimental projects and solutions and help reduce or logically deduct conclusions
(Classification, Regression, Ranking, Clustering, Dimensionality)