Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
1© 2015 The MathWorks, Inc.
Get More From Your Data
with Data Analytics
Francesca Perino
2
3
4
5
Buildings have thermodynamic properties
𝜕𝑢
𝜕𝑡− 𝛼
𝜕2𝑢
𝜕𝑥2+𝜕2𝑢
𝜕𝑦2+𝜕2𝑢
𝜕𝑧2= 0
6
Temperatures change
7
Electricity demand varies
8
Humans have comfort bounds
9
𝜕𝑢
𝜕𝑡− 𝛼
𝜕2𝑢
𝜕𝑥2+𝜕2𝑢
𝜕𝑦2+𝜕2𝑢
𝜕𝑧2= 0
10
11
1. Diverse and/or Big Data
2. Advanced Algorithms
3. Deployment
Traits of Data Analytics applications
12
Why MATLAB?
Develop embedded
systems with analytics
powered functionality
3
Enable Domain
Experts to be
Data Scientists
2
Analytics that increasingly require both business
and engineering data
1
Data Scientist
Embedded Systems
DATA
• Engineering, Scientific, and Field
• Business and Transactional
Business Systems
Develop analytics to run on both enterprise and
embedded platforms
4
13
Why MATLAB?
Develop embedded
systems with analytics
powered functionality
3
Enable Domain
Experts to be
Data Scientists
2
Data Scientist
Smarter Embedded Systems
DATA
• Engineering, Scientific, and Field
• Business and Transactional
Business Systems
Develop analytics to run on both enterprise and
embedded platforms
4
Analytics that increasingly require both business
and engineering data
1
14
Business and Engineering Data
“No matter what industry our client is in, and no matter what data they ask us to analyze—text,
audio, images, or video—MATLAB enables us to provide clear results faster.”
Dr. G Subrahmanya VRK Rao, Cognizant
File I/O
• Text
• Spreadsheet
• XML
• CDF/HDF
• Image
• Audio
• Video
• Geospatial
Repositories
• Databases (SQL)
• NoSQL
• Hadoop
File I/O
• Text
• Spreadsheet
• XML
Business and Transactional Data Engineering, Scientific, and Field Data
Web Sources
• HTML
• Mapping
• Financial datafeeds
• RESTful
• JSON
Communication Protocols
• CAN (Controller Area Network)
• DDS (Data Distribution Service)
• OPC (OLE for Process Control)
• XCP (eXplicit Control Protocol)
Real-Time Sources
• Sensors
• GPS
• Instrumentation
• Cameras
• Communication systems
• Machines (embedded systems)
15
Accessing Data
16
Datastore
HDFS
Reduce
Node
Node
Node Data
Data
Data
Map
ReduceMap
ReduceMap
Map Reduce
Map
Map
Reduce
Reduce
17
Table
18
Preprocessing Data
19
Signal Processing
cheby2
filter
rms
pwelch
periodogram
xcov
findpeaks
movmean
movstd
…
20
Image Processing
Perform image processing, analysis, visualization, and algorithm development
Image display and exploration
Image enhancement
Image analysis
Morphological operations
Image registration
Geometric transformation
ROI-based processing
21
Feature Engineering – Extracting Information from Data
Data type Common Techniques for Deriving Features
Sensor data Peak analysis
Pulse and transition metrics
Spectral measurements (power, bandwidth, mean frequency,
median frequency)
Image and video data Bag of visual words
HOG (Histogram of Oriented Gradients)
Minimum Eigenvalue algorithm
Local feature descriptors
Edge detection
Transactional data Decomposing timestamps into components (day, month, day of
week, etc.)
Calculation of aggregate values
22
Why MATLAB?
Develop embedded
systems with analytics
powered functionality
3
Analytics that increasingly require both business
and engineering data
1
Data Scientist
Smarter Embedded Systems
DATA
• Engineering, Scientific, and Field
• Business and Transactional
Business Systems
Develop analytics to run on both enterprise and
embedded platforms
4
Enable Domain
Experts to be
Data Scientists
2
23
Enabling Domain Experts to be Data Scientists
Symbolic
Computing
Neural
Networks
OptimizationSignal
Processing
Image
Processing
Control
Systems Financial
Modeling
Apps Language
Machine
LearningStatistics
“MATLAB has helped accelerate our R&D and deployment with its robust numerical algorithms, extensive visualization and analytics tools, reliable optimization routines, support for object-oriented programming, and
ability to run in the cloud with our production Java applications.”Borislav Savkovic, BuildingIQ
24
Apps - Classification Learner app
25
Redesigned execution engine runs MATLAB code faster
– All MATLAB code can be JIT compiled
– A platform for future improvements
Language - MATLAB Execution Engine
The examples of all FSDA functions have been monitored under R2012a,
R2014b, R2015a, R2015b. Results seem to indicate that, generally, on
computationally intensive mathematical/statistical function R2015b yields
remarkable time improvements.
26
Why MATLAB?
Enable Domain
Experts to be
Data Scientists
2
Analytics that increasingly require both business
and engineering data
1
Data Scientist
Smarter Embedded Systems
DATA
• Engineering, Scientific, and Field
• Business and Transactional
Business Systems
Develop analytics to run on both enterprise and
embedded platforms
4Develop embedded
systems with analytics
powered functionality
3
27
Smarter Embedded Systems
INTEGRATION
IMPLEMENTATION
DESIGN
TE
ST
AN
D V
ER
IFIC
AT
ION
RESEARCHREQUIREMENT
S
MCU DSP FPGA ASIC
VHDL,
VerilogC, C++
Environment Models
Physical Components
Algorithms
Structured
Text
PLC
GM
Climate control
Airbus
Battery management
Daimler
Cruise controller
manroland
Printing presses
Sonova
Hearing implantsWeinmann
Transport ventilator
Festo
Industrial robots
FLIR
Thermal imaging
ABB
Smart Grid controller
28
MATLAB Code Generation
29
C Code Generation
30
Why MATLAB?
Develop embedded
systems with analytics
powered functionality
3
Enable Domain
Experts to be
Data Scientists
2
Analytics that increasingly require both business
and engineering data
1
Data Scientist
Smarter Embedded Systems
DATA
• Engineering, Scientific, and Field
• Business and Transactional
Business Systems
Develop analytics to run on both enterprise and
embedded platforms
4
31
Where does the processing happen?
Devices Business Systems
Visibility Self, Maybe Neighbors Entire System
Power Battery (Low) Grid (Unlimited)
Processing Microprocessor (Low) CPU’s (Unlimited)
Storage Memory (Minimum Buffer) HD’s (Unlimited)
Updates Difficult Easy
32
Why not transfer all the data?
1. Data privacy concerns
2. Cost of network/storage
3. Power required to transmit data from device (for wireless)
4. Response time
33
Splitting computation
34
Smart Systems: Wearable Healthcare Technology
35
MATLAB
ExcelAdd-in
Hadoop
MATLAB
Compiler
StandaloneApplication
Deploying Algorithms to Enterprise Systems
MATLAB Compiler enables sharing MATLAB programs without integration programming
MATLAB Compiler SDK provides implementation and platform flexibility for software developers
MATLAB Production Server provides the most efficient development path for secure and
scalable web and enterprise applications
MATLAB
Compiler SDK
C/C++
Java
.NET
MATLABProduction
Server
Python
36
Enterprise Integration – Forecasting Model
37
MATLAB Differentiators
Develop embedded
systems with analytics
powered functionality
3
Enable Domain
Experts to be
Data Scientists
2
Analytics that increasingly require both business
and engineering data
1
Data Scientist
Smarter Embedded Systems
DATA
• Engineering, Scientific, and Field
• Business and Transactional
Business Systems
Develop analytics to run on both enterprise and
embedded platforms
4
38
Learn More