Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Social Life Networks:
EventShop and Personal Event Shop
Experiential Systems Laboratory
University of California, Irvine
S...
Outline
• Social Life Networks
• EventShop
• Personal EventShop
• Predictive Analytics
Fundamental Problem
Web 1.0
Connecting People to Documents
Web 2.0
Connecting People to People
“Social Life Network”
Conne...
EventShop : Global Situation Detection
Situation
Recognition
Evolving Global Situation
Personal
Situation
Recognition
Pers...
EventShop: Recognizing Situations
from Heterogeneous Data Streams
Siripen Pongpaichet
spongpai@uci.edu
Big Data “NEED” Big Insight
Figure Ref: http://damfoundation.org/2012/09/big-data-big-deal/
Related Services
7
http://google.org/crisismap/sandy-2012
Mash Up: Google Crisis Maps
http://google.org/crisismap/2013-utt...
“EventShop” towards SLN
7/03/2013 8
[Pongpaichet 2013] EventShop: recognizing situations in web data streams
From Heterogeneous Data Sources
to Situation Recognition
7/03/2013 9
Example Notification / Alerts:
You are currently in t...
Data Unified Format: STT
• Each data source is transformed into unified
data format, S-T-T (Space - Time - Theme)
• Then e...
E-mage Algebra & Operators

Pattern Matching
Aggregatioin

@ Characterization
∏ Filter
 Segmentation
72%
+
+
Growth Rat...
EventShop Web UI
Multi-Spatio-Temporal
Bounding Boxes and Resolutions
• “Pyramid of E-mage” resolution
is introduced to represent the
real ...
Rasterization and Error Propagation
• Data Error Factors:
– Uncertainty of data stream
– Data loss during data aggregation...
Personal Event Shop:
Recognizing Evolving Situation of Person
Laleh Jalali
lalehjlal@uci.edu
Personal EventShop
• Understanding a person
• Personicle: Personal Chronicle
• Detecting evolving personal situation
• Rec...
Wearable 2013:
Data Data Everywhere, nor any Insight to Use.
http://www.medhelp.org/
HEALTH
PERSONA
Logical Sensor
Physical Sensors
Life
Event
Motion
Event
Physiological
Event
Food
Event
Personicle
Health Pe...
Event Categories
• Event Definition:
▫ NOT abnormal observation.
▫ NOT point event.
▫ BUT interval event
20
Motion event S...
Input Data
Manager
Raw Data Streams
Event Streams** Predictive
Data
Analytics
Personal Data Warehouse
PersonalDataSources
...
Activity Detection
Location
Data Stream
Accelerometer
Data Stream
Feature Extraction
Learning and Inference
Data Collectio...
Activity Level
Asthma Triggers
Indoor Allergens
Outdoor Allergens
Smoke
Air Pollution
Chemical Irritants
Predictive Techniques for
Preventive Advise
25
Insight:
Exercise Triggers My Asthma
• Red: “Severe asthma attack”.
• Orange: High activity level.
• Exercise triggers my ...
Predictive Analytics:
Personalized Asthma Risk Prediction
Mengfan Tang
mengfant@uci.edu
Enrich Personalized Asthma Risk
• Predict air quality at air quality measuring
sites.
• Interpolate air quality at the loc...
Daily Ozone Data
Ref- http://www.arb.ca.gov/aqmis2/aqmis2.php
Time-Series Prediction
Autoregressive model:
Xt = ji Xt-i
i=1
p
å +et
K-Nearest Neighbors Regression
Air Quality Index
A limited number of sites.
How can we predict the air quality in the area not
covered by the sites?
Historical air quality data
Interpolate Air Quality Index
The Learned models
trained on multiple data
sources
Learned models
Interpolate Air Quality Index
Personalized Asthma Risk Prediction
• The predictive model
– Correlation with Personicle
• Temporal correlation in a locat...
Upcoming SlideShare
Loading in …5
×

Social Life Networks (Eventshop and Personal Event Shop)

236 views

Published on

presented at USC workshop on Web Observatory

Published in: Data & Analytics
  • Be the first to comment

  • Be the first to like this

Social Life Networks (Eventshop and Personal Event Shop)

  1. 1. Social Life Networks: EventShop and Personal Event Shop Experiential Systems Laboratory University of California, Irvine Siripen Pongpaichet, Laleh Jalali, Mengfan Tang (Adviser: Ramesh Jain)
  2. 2. Outline • Social Life Networks • EventShop • Personal EventShop • Predictive Analytics
  3. 3. Fundamental Problem Web 1.0 Connecting People to Documents Web 2.0 Connecting People to People “Social Life Network” Connecting Needs to Resources Effectively, Efficiently, and Promptly In given situations.
  4. 4. EventShop : Global Situation Detection Situation Recognition Evolving Global Situation Personal Situation Recognition Personal EventShop Evolving Personal Situation Need- Resource Matcher Recommendation Engine PersonaDatabase Resources Needs Data Ingestion Wearable Sensors Calendar Location…. DataSources …. Data Ingestion and aggregation Database Systems Satellite Environmental Sensor Devices Social Network Internet of Things Actionable Information
  5. 5. EventShop: Recognizing Situations from Heterogeneous Data Streams Siripen Pongpaichet spongpai@uci.edu
  6. 6. Big Data “NEED” Big Insight Figure Ref: http://damfoundation.org/2012/09/big-data-big-deal/
  7. 7. Related Services 7 http://google.org/crisismap/sandy-2012 Mash Up: Google Crisis Maps http://google.org/crisismap/2013-uttrakhand-floods one-touch SOS Mobile Applications
  8. 8. “EventShop” towards SLN 7/03/2013 8 [Pongpaichet 2013] EventShop: recognizing situations in web data streams
  9. 9. From Heterogeneous Data Sources to Situation Recognition 7/03/2013 9 Example Notification / Alerts: You are currently in the area where there is a high chance of flooding, these are available shelters within 10 miles around you. Space Time Situation Resources People (Space, Time, Theme) [Pongpaichet 2013] EventShop: recognizing situations in web data streams
  10. 10. Data Unified Format: STT • Each data source is transformed into unified data format, S-T-T (Space - Time - Theme) • Then each STT data stream is aggregated to form E-mage (Event-Image) stream “iphone” popularity in USA from Tweets data source
  11. 11. E-mage Algebra & Operators  Pattern Matching Aggregatioin  @ Characterization ∏ Filter  Segmentation 72% + + Growth Rate = 125% Data Supporting parameter(s) OutputOperator Type + Segmentation methods Property required Pattern Mask 7/03/2013 12
  12. 12. EventShop Web UI
  13. 13. Multi-Spatio-Temporal Bounding Boxes and Resolutions • “Pyramid of E-mage” resolution is introduced to represent the real world in E-mage at different (zoom) levels. • Each Stel (a pixel in the E-mage) represents a single fixed ground location. • Precision vs Computational Cost
  14. 14. Rasterization and Error Propagation • Data Error Factors: – Uncertainty of data stream – Data loss during data aggregation – Uncertainty during data conversion – Data error during data conversion • To design the most optimum situation recognition model, we need to find the new cost evaluation method that will consider both data accuracy and computational cost.
  15. 15. Personal Event Shop: Recognizing Evolving Situation of Person Laleh Jalali lalehjlal@uci.edu
  16. 16. Personal EventShop • Understanding a person • Personicle: Personal Chronicle • Detecting evolving personal situation • Recommending action
  17. 17. Wearable 2013: Data Data Everywhere, nor any Insight to Use. http://www.medhelp.org/
  18. 18. HEALTH PERSONA Logical Sensor Physical Sensors Life Event Motion Event Physiological Event Food Event Personicle Health Persona Framework: Humans as Actuators. 19 Environmental Sensors
  19. 19. Event Categories • Event Definition: ▫ NOT abnormal observation. ▫ NOT point event. ▫ BUT interval event 20 Motion event Stream Physiological event Stream Personicle
  20. 20. Input Data Manager Raw Data Streams Event Streams** Predictive Data Analytics Personal Data Warehouse PersonalDataSources * location stream, activity stream, motion stream, calendar stream ** life-event stream, motion-event stream, physiological-event stream, food-event stream Personal EventShop Platform Data Streams * Data Analytics Persona Profile Info. Asthma History Domain knowledge Asthma Attack Symptoms & Severity Recom. Engine Evolving personal situation Moves app Nike Fuel Foursquare Google Calendar Personicle Actionable Information 21 Environmental Factors
  21. 21. Activity Detection Location Data Stream Accelerometer Data Stream Feature Extraction Learning and Inference Data Collection Extracted Feature Set Recognition Model • Frequency Domain • Energy • Entropy • Coefficients Sum • DC Component • Time Domain • Mean , Median • Std Deviation • Min, Max, Range • Average Absolute Difference • Average Resultant Variation
  22. 22. Activity Level
  23. 23. Asthma Triggers Indoor Allergens Outdoor Allergens Smoke Air Pollution Chemical Irritants
  24. 24. Predictive Techniques for Preventive Advise 25
  25. 25. Insight: Exercise Triggers My Asthma • Red: “Severe asthma attack”. • Orange: High activity level. • Exercise triggers my asthma! Personicle Activity Stream t1 t2 t3 t4 t5
  26. 26. Predictive Analytics: Personalized Asthma Risk Prediction Mengfan Tang mengfant@uci.edu
  27. 27. Enrich Personalized Asthma Risk • Predict air quality at air quality measuring sites. • Interpolate air quality at the locations not covered by measuring sites. • Predict personalized asthma risk by using EventShop and Personal EventShop.
  28. 28. Daily Ozone Data Ref- http://www.arb.ca.gov/aqmis2/aqmis2.php
  29. 29. Time-Series Prediction Autoregressive model: Xt = ji Xt-i i=1 p å +et
  30. 30. K-Nearest Neighbors Regression
  31. 31. Air Quality Index A limited number of sites. How can we predict the air quality in the area not covered by the sites?
  32. 32. Historical air quality data Interpolate Air Quality Index The Learned models trained on multiple data sources
  33. 33. Learned models Interpolate Air Quality Index
  34. 34. Personalized Asthma Risk Prediction • The predictive model – Correlation with Personicle • Temporal correlation in a location • Geo-correlation between locations – Two sets of features • Spatially-related • Temporally-related – Machine learning methods • Supervised learning

×