Digitally sensing people-place interactions in near real-time using social, mobile and IoT data. Talk delivered for Brainpool cognitive and data science group in London, 1 December 2017.
Delhi 99530 vip 56974 Genuine Escort Service Call Girls in Kishangarh
Urban Cognitive Analytics
1. Urban Cognitive Analytics
Sensing people-place interactions
from urban digital footprints
Doctoral Researcher: Sharon Richardson
Supervised by: Professor Andrew Hudson-Smith and Dr Hannah Fry
sharonr@joiningdots.com | @joiningdots | linkedin.com/in/sharonr
3. Context - Presence - Behaviour
What’s happening? Where is everybody? What are they doing?
weekday weekend
4. Urban Digital Footprints
Daily volumes
from May to
August 2016
Contextual
factors:
• Seasonal expectations
• Workday vs holiday
• Climate (temperature
and precipitation)
• Situation (events and
social behaviours)
• Space and time
5. Linguistic Variation in Tweets
Sat June 4th
Cosine Similarity Sat Jun 4th Sat Jun 11th Sat Jun 18th Sat Jun 25th
Sat Jun 4th 1.0000 0.2360 0.1176 0.2365
Sat Jun 11th 0.2360 1.0000 0.4082 0.7092
Sat Jun 18th 0.1176 0.4082 1.0000 0.3904
Sat Jun 25th 0.2365 0.7092 0.3904 1.0000
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Sat 4th
Sat 11th
Sat 18th
Sat 25th
Similarity in terms used between dates
Sat 4th Sat 11th Sat 18th Sat 25th
Events at the Queen Elizabeth Olympic Park:
- AC/DC concert at stadium on Saturday 4th June
- International hockey tournament at VeloHTC from 10th to 26th June
Sat June 11th Sat June 18th Sat June 25th
6. Emotion Scores
0
0.005
0.01
0.015
0.02
0.025
affec+on anger anxiety expressive glory posi+ve sadness social
Twitter RID Emotion Scores
07/08/16
21/08/16
14/08/16
28/08/16
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
A3_HackneyWick B4_Stadium C3_Westfield C4_Aquatics
Average sentiment score (+ve/-ve words) for tweets
7th Aug
21st Aug
14th Aug
28th Aug
running under the a at hackney wick dark and a little bit sinister running summer london
louisregan hopefully man city will demolish your shit team today put the london stadium in its place coyidiots
c c rail can you negotiate with greateranglia tflrail to still send c c rail into stratford even they all terminate that station
jam n ib rico con queso jam n queso jamon cheese lunch piecesofheaven enriquethomas
that awkward moment when you help your parents miss their train to holland riasmit rs
last street beat battle trailer don t miss the crazy battles friday th september all
i actually despise everything westham have become the move to the london stadium has ripped the soul out of that club
hoaxerttv mrblakerz poseyxr elevatemunoz orbit dramis i randomly hopped on gb this am and there was one today
the olympic stadium on a sunday morning is without a doubt the weirdest place in london
have taken the perhaps terrible decision to go to westfield in stratford to look for shoes probably going to be hellishly packed
kicchasudeep hi kiccha bro iam already watch k twise and now i am watching with my brother now intermission orbit break
on my way to gatwickairport i m going to miss her so much here she is fighting her favourite
back at one of my career highlights but also where i nearly died at lee valley velopark
santandercycles very annoying to reach end of cs stratford and discover that there s no docking station near there extra to find one
telefootball can see the move to to the london stadium being a curse for hammers
sunday behaviour at worst with food foodpron at westfield stratford shopping centre london
why do i frown at westfield stratford city
humans put nasajuno in polar orbit around jupiter human begins fb post i am disappointed with color
at shake shack stratford
darnz baiting out paedophiles in stratford westfield gone wrong got bottled jyqcvxjqyv
Events at the Queen Elizabeth Olympic Park:
- Football at the London Stadium on 7th and 14th
- No scheduled events on 21st and 28th
- Incident on the 28th???
Term analysis (punctuation removed before tokenisation)
19. Community Detection
Community Detection Scores
fg = Fast Greedy
gn = Girvan-Newman (baseline)
im = Infomap
lv = Louvain
pl = Pons and Latapy
g = simplified graph (remove loops and multiples)
g1 = original graph converted from Gephi to iGraph
g1.gexf = original graph as imported (Gephi file)
20. Self-supervised learning
Step 1: Analyse (current) network
- Detect communities forming within network
- Convert each community into a set of terms
Step 2: Create dictionary
- Merge sets with <4 items into single ‘noise’ set
- Create dictionary of topics with hierarchy for topic / term set
Step 3. Classify tweets
- calculate tweet score based on similarity with each topic
- classify tweet with topic that scores the highest similarity
Step 4: Evaluate clusters
34
26
20
9 9
8
7 7 7
4 4
3 3 3 3
2 2 2 2 2 2 2 2 2 2 2 2
0
5
10
15
20
25
30
35
40
27 8 25 5 14 1 3 9 13 4 18 2 6 19 22 7 10 11 12 15 16 17 20 21 23 24 26
Number of Hashtags per Community (27 communities)
Number of Hashtags
21. The science of where people are
image: https://pxhere.com/en/photo/1392001
Sensing population dynamics:
• Administrative
Residential census
• Ambient
Average workday
• Active
Near real-time
Quantifying the population at risk…
Anticipating behaviour:
levels of spatial familiarity
What’s around the corner?
22. Urban Cognitive Analytics
Sensing people-place interactions
from urban digital footprints
Doctoral Researcher: Sharon Richardson
Supervised by: Professor Andrew Hudson-Smith and Dr Hannah Fry
sharonr@joiningdots.com | @joiningdots | linkedin.com/in/sharonr