12
Björn Bjurling, [email protected] Daniel Gillblad, [email protected] Anders Holst, [email protected] Swedish Institute of Computer Science BIG DATA AND ANALYTICS

BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

Björn Bjurling, [email protected] Daniel Gillblad, [email protected] Anders Holst, [email protected]

Swedish Institute of Computer Science

BIG DATA AND ANALYTICS

Page 2: BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

www.sics.se

AGENDA

•  What is big data and analytics?

•  and why one must bother

•  Examples of big data for vehicles

•  Summary and take away lessons

Page 3: BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

www.sics.se

BACKGROUND: PARADIGM SHIFTS •  Advances in hardware and computer

systems •  Cheaper storage, faster CPUs, and

faster networking •  Parallel computing, Cloud computing

•  Abundance of data •  Sensor systems revolution •  Internet services, Social media •  Mobility and connectedness •  Improved Data collection capabilities

•  Data analysis •  Scale and complexity

enable/require new algorithms •  Success stories

•  Facebook, Google, …

Page 4: BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

www.sics.se

BIG DATA: MANY CHALLENGES

•  Computations and platforms •  Hardware / Infrastructure / Data Centers •  Storage/communication/networking •  Programming concepts •  Code/ Compilation/ Scheduling

•  Algorithms •  Scalability •  Complexity •  Decentralization •  Time requirements

•  Data analysis •  Representation / Modelling •  Domain knowledge •  Visualization

•  Deployment •  Business models / Services •  Security / Privacy / Legal aspects •  Power / Environment

Page 5: BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

www.sics.se

Data cleaning

Representation

Validation

Deployment

Neural Networks

Logical Inference

Case-based

Statistical Methods

(BIG) DATA ANALYTICS IN PRACTICE

Page 6: BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

www.sics.se

BIG DATA PROMISES •  Extraction of valuable information from

large data sets

•  Increasing volumes of data lead to increasing value of extracted information

•  Uncovering of otherwise hidden and valuable information

•  Connected vehicles + big data analytics

•  Novel services

•  Improved efficiency and productivity

•  Competitive edge

Page 7: BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

www.sics.se

TRENDS IN BIG DATA RESEARCH

•  Strategies for surviving the data flood

•  Learning Representation

•  Taking advantage of structure:

•  Graph Processing

•  Big data transformed to Small data

•  Platform/algorithm interplay

•  Local vs global computation

•  Streaming data

•  Store and communicate models

Page 8: BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

www.sics.se

REAL TIME TRAFFIC AWARENESS

•  High availability of traffic reports and collection of vehicle-based positioning data

•  Allows modelling and prediction of traffic situation for individual vehicles

•  Toyota will launch its Big data traffic information system for providing services for •  optimal routes •  predictions of travelling times

Page 9: BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

www.sics.se

FLEET MANAGEMENT

•  ARI Fleet collect thousands of data types from each vehicle in its fleet (a million vehicles)

•  Applying state-of-the-art big data analytics helps ARI Fleet make substantial savings through •  timely and precise

maintenance scheduling

•  improved transport scheduling

Page 10: BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

www.sics.se

MANUFACTURING

•  Collecting and anlysing data from

•  Driver behavior

•  Vehicle behavior

•  Service and maintenance cycles

•  … can give manufacturers valuable insights into how to improve driving experience and security aspects already in design stage

•  Range Rover’s Best Suv of the year (2012) model Evoque was designed taking into account extensive simulations based on analysis of collected data from the performance and behavior of earlier models

Page 11: BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

www.sics.se

Res

ourc

e m

anag

emen

t

THE DATA DRIVEN SYSTEMS STACK, EXAMPLES OF WORK AT SICS

Stream processing,

Pig

Information Centric Networking, SDN

HOPS as Platform As a Service

Scalable HDFS

SicsthSense Network Search SDN Monitoring

Text and Social Media

MapReduce Stratosphere Spark

Autonomous RAN

Ja Be Ja Graph

Clustering

Anomaly / change

detection

Traffic and mobility

modeling

Data collection

Networking

Storage

Computing

Frameworks

Domain specific

Page 12: BIG DATA AND ANALYTICS · • Big data analytics does not come out of the box • Need domain knowledge for meaningful data analysis • Every domain of application of BDA requires

www.sics.se

TAKE AWAY LESSONS

•  Big data analytics does not come out of the box

•  Need domain knowledge for meaningful data analysis

•  Every domain of application of BDA requires unique analysis, modelling, and deployment

•  Paradigm change in ICT and Society

1.  Information is power – extracting value from data is becoming the crucial competitive advantage

2.  ICT is becoming data and service centric – Application driven; compute, storage and communication viewed as services

3.  ICT is becoming an integrated part of products and services