Topology-Guided Vis isual Exploratory ry Analysis Harish - - PowerPoint PPT Presentation

▶
topology guided
SMART_READER_LITE
LIVE PREVIEW

Topology-Guided Vis isual Exploratory ry Analysis Harish - - PowerPoint PPT Presentation

Jul 07, 2023 464 likes •865 views

Topology-Guided Vis isual Exploratory ry Analysis Harish Doraiswamy NYU Center for Data Science Scan All Fish 2 Scan All Fish 3 TopoAngler 4 Data Exhaust from Cities 5 Data Exhaust from Cities Infrastructure Environment People 6

slide-1
SLIDE 1

Topology-Guided Vis isual Exploratory ry Analysis

Harish Doraiswamy

NYU Center for Data Science

slide-2
SLIDE 2

Scan All Fish

2

slide-3
SLIDE 3

Scan All Fish

3

slide-4
SLIDE 4

TopoAngler

4

slide-5
SLIDE 5

Data Exhaust from Cities

5

slide-6
SLIDE 6

Data Exhaust from Cities

Infrastructure Environment People

6

slide-7
SLIDE 7

Open Urban Data

7

slide-8
SLIDE 8
  • Yellow cab trips
  • ~175 million trips / year
  • Spatial-Temporal
  • 2 spatial attributes
  • 2 temporal attributes
  • Other attributes
  • Fare, tip
  • Distance
  • Duration
  • …

NYC Taxi Data

8

slide-9
SLIDE 9

2011 2012

9

slide-10
SLIDE 10

Idea: Use Topology of the Data

10

slide-11
SLIDE 11

Idea: Use Topology of the Data

11

slide-12
SLIDE 12

Exploring Features

  • Several features per time step
  • Group similar features within a larger time interval
  • Represents “macro” events
  • Similarity
  • Graph similarity: Shape
  • Persistence / Volume: Topological similarity
  • Key for each group
  • Average shape and volume
  • Efficient search

12

slide-13
SLIDE 13

Guiding Users towards Interesting Events

  • Properties of Macro Events

Two micro-events that occur on consecutive hours Two micro-events that occur on consecutive days An event that occurs every hour during a week Two micro-events that occur on consecutive weeks Rare occurence of these events Frequent

  • ccurance

13

slide-14
SLIDE 14

Rare and Interesting Features - Hourly

  • October

Halloween Parade

14

slide-15
SLIDE 15

Daily

  • October
  • 1. Hispanic Day Parade (Oct 9 2011)
  • 2. Columbus Day Parade (Oct 10 2011)

15

slide-16
SLIDE 16

Weekly

  • August
  • No. of weeks = 3
  • NYC Summer Streets

16

slide-17
SLIDE 17

5 Borough Bike Tour 2011 (1 May 2011) Query

5 Borough Bike Tour 2012 (6 May 2012) Dominican Day Parade 2011 (14 August 2011) Dominican Day Parade 2012 (12 August 2012) Gaza Solidarity Protest NYC (18 November 2012)17

slide-18
SLIDE 18

Frequent Features

  • Taxi hotspots
  • Filter over time

General trends Night time trends

18

slide-19
SLIDE 19

Frequent Features

  • Maxima: Taxi hotspots
  • Filter over time

General trends Night time trends

Using Topological Analysis to Support Event-Guided Exploration in Urban Data, TVCG 2014.

19

slide-20
SLIDE 20
  • Design of public spaces
  • Understand what works / doesn’t work in one city
  • Use this to improve design in another city
  • Understand properties of neighborhoods
  • Compare “activity” between neighborhoods with similar

properties

  • Compare properties between neighborhoods with similar

“activity”

How to compare cities?

20

slide-21
SLIDE 21
  • How do cities behave during different times?
  • Summer vs. Winter
  • Weekdays vs. Weekends
  • Data sets about different cultural communities in a

city

  • What patterns do the different communities follow?
  • How do these patterns compare?

How to analyze / compare different properties

  • f a city?

21

slide-22
SLIDE 22

Urban Pulse

  • Flickr activity in New York City

7:00 am

1

22

slide-23
SLIDE 23

Urban Pulse

  • Flickr activity in New York City

7:00 am

1

11:00 am

23

slide-24
SLIDE 24

Urban Pulse

  • Flickr activity in New York City

7:00 am

1

7:00 pm

24

slide-25
SLIDE 25

Urban Pulse

  • Flickr activity in New York City

7:00 am

1

11:00 pm

25

slide-26
SLIDE 26

Temporal Resolutions

  • Compute functions along 3 resolutions

Time of Day Day of Week Month of Year

26

slide-27
SLIDE 27

Step 1: Identify Pulse Locations

  • Set of scalar functions over time
  • Identify all maxima
  • Location of prominent pulses
  • is a high persistent maxima in at least

1 time step

  • is a high persistent maxima in at least

1 resolution

  • 1. Identify Locations
  • 2. Quantify Pulse

27

slide-28
SLIDE 28

Step 2: Quantifying Pulse

  • 3 Beats to quantify the pulse at each location
  • Significant Beats
  • Is the location a high persistent maximum?

Bs

  • 1. Identify Locations
  • 2. Quantify Pulse

28

slide-29
SLIDE 29

Step 2: Quantifying Pulse

  • 3 Beats to quantify the pulse at each location
  • Maxima Beats
  • Is the location a maximum?

Bs Bm

  • 1. Identify Locations
  • 2. Quantify Pulse

29

slide-30
SLIDE 30

Step 2: Quantifying Pulse

  • 3 Beats to quantify the pulse at each location
  • Function Beats Bf
  • Variation of the function values

Bs Bm Bf

  • 1. Identify Locations
  • 2. Quantify Pulse

30

slide-31
SLIDE 31

Step 2: Quantifying Pulse

Time of Day Day of Week Month of Year

  • 1. Identify Locations
  • 2. Quantify Pulse

31

slide-32
SLIDE 32

Step 2: Quantifying Pulse

B1 B2 B3 B4 B5 B6 B7 B8 B9

Rank Signature Data Oblivious

  • 1. Identify Locations
  • 2. Quantify Pulse

32

slide-33
SLIDE 33

Step 2: Quantifying Pulse

Compare Signature Data Oblivious

  • 1. Identify Locations
  • 2. Quantify Pulse

B1 B2 B3 B4 B5 B6 B7 B8 B9

33

slide-34
SLIDE 34

Urban Pulse Interface

34

slide-35
SLIDE 35

Use Case: Understanding Public Spaces

  • Typically classified together as being similar

Bryant Park Union Square Rockefeller Center

35

slide-36
SLIDE 36

Use Case: Understanding Public Spaces

Bryant Park Union Square Rockefeller Center

36

slide-37
SLIDE 37
  • Yellow cab trips
  • ~175 million trips / year
  • Spatial-Temporal
  • 2 spatial attributes
  • 2 temporal attributes
  • Other attributes
  • Fare, tip
  • Distance
  • Duration
  • …

NYC Taxi Data

37

slide-38
SLIDE 38

Work done together with Alex Bock, Theodoros Damoulas, Nivan Ferreira, Juliana Freire, Bruno Gonçalves, Mondrian Hsieh, Marcos Lage, Fabio Miranda, Claudio Silva, Adam Summers, Luc Wilson, Kai Zhao

38

Topology-Guided Vis isual Exploratory ry Analysis

https://www.github.com/harishd10 https://github.com/ViDA-NYU