[PPT] - S Space-in-Time and Time-in-Space Self-Organizing Maps i Ti d Ti PowerPoint Presentation

SLIDE 1

S i Ti d Ti i S S lf O i i M Space-in-Time and Time-in-Space Self-Organizing Maps for Exploring Spatiotemporal Patterns

Gennady Andrienko & Natalia Andrienko htt // l ti t http://geoanalytics.net Sebastian Bremm, Tobias Schreck, Tatiana von Landesberger TU Darmstadt Peter Bak, Daniel Keim

Univ. Konstanz

DFG Priority Research Program on Visual Analytics SPP 1335

10 June 2010, presentation at EuroVis

Bordeaus, FR

DFG Priority Research Program on Visual Analytics, SPP 1335

SLIDE 2

S ti l Ti S i D t St t Spatial Time Series: Data Structure

Spatial references: states of the USA Spatial references: states of the USA Temporal references: years from 1960 till 2000 (41) Attributes: population + various crime rates Attributes: population + various crime rates

2

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 3

S ti l Vi li ti i t d di Spatial Visualizations: animated maps, diagram maps

3

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 4

T l i li ti Temporal visualizations

4

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 5

S l bilit bl Scalability problem

What if we have
Multiple attributes
Many places

L ti i

Long time series
Interactive visualization is not sufficient
We need grouping in space and time => Clustering

5

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 6

D t t Data sets

Cars in Milan, Italy
175,890 trajectories of 17,241 cars over 7 days
2,075,216 records <id,x,y,t,speed>

A t d 18 22 1k

2

t l i

Aggregated over 18x22 1km2 rectangular regions

and 7x24 hourly intervals

USA crime statistics
7 crime attributes
52 states
52 states
41 years

6

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 7

A h Approach

7

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 8

A h Approach

2) Group time intervals by similarity of spatial situations: clustering “Space in Time” 1) Group places by ) G oup p aces by similarity of temporal dynamics: l t i clustering “Time in Space”

8

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 9

SOM SOM

Self-Organizing Map (Kohonen 2001) is a neural network type vector projection and

quantization algorithm.

By means of a competitive, iterative training process, a network of prototype vectors

(or neurons, or cells) is trained (adjusted) to the input vector data. ( ) ( j ) p

The output of the algorithm is a network of vectors that is approximately topology

preserving w.r.t. the input data. Th t k b i t t d t f l t d i lt l t

The network can be interpreted as a set of clusters and simultaneously as a map to

lay out the input data elements (e.g., in the nearest neighbor sense w.r.t. the prototypes).

Typically, two-dimensional rectangular or hexagonal prototype vector networks are

assumed.

The capability of SOM to arrange input data in a regular network structure provides

good opportunities for visualization.

9

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 10

S i Ti d Ti i S SOM i li ti Space-in-Time and Time-in-Space SOMs: visualization

1.

Bars on top of a cell show number of objects inside

2.

Shading of borders between cells reflects similarity of features

3.

Similarity of colors also reflects similarity of the features C l j t d th

4.

Colors are projected on other displays: maps (if grouping places) and and time graphs (if grouping time intervals)

10

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 11

Ti i S SOM f d i i d Time-in-Space SOM of driving speeds

11

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 12

Ti i S SOM f d i i d Time-in-Space SOM of driving speeds

Inside cells:

index images (what is grouped)

12

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 13

Ti i S SOM f d i i d Time-in-Space SOM of driving speeds

Inside cells:

feature images (what are the features)

13

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 14

Ti i S SOM f d i i d Time-in-Space SOM of driving speeds

Inside cells:

index images (what is grouped) feature images (what are the features)

14

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 15

Ti i S SOM d t il Time-in-Space SOM: details

15

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 16

Ti i S SOM d t il Time-in-Space SOM: details

16

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 17

S i ti SOM (i d i ) Space-in-time SOM (index images)

17

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 18

S i ti SOM (i d & f t i ) Space-in-time SOM (index & feature images)

18

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 19

S i ti SOM ( l f ti i t l ) Space-in-time SOM (colours of time intervals)

19

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 20

S i ti SOM ( l f ti i t l ) Space-in-time SOM (colours of time intervals)

20

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 21

D t t th t d Detect the expected

One more data set about Milan:

mobile phone calls for 9 days aggregated by hours and regions

We expect high periodicity and clear regionalization of the calling activity

21

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 22

Ti i S Time-in-Space SOM of mobile phone calls

22

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 23

S i Ti Space-in-Time SOM of mobile phone calls

23

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 24

Di th t d Discover the unexpected

USA crime statistics
7 crime attributes
52 states

41

41 years

Problems Problems:

Different ranges of attributes
Outliers

Outliers

24

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 25

C i tt ib t Crime attributes

25

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 26

C i tt ib t li d Crime attributes: normalized

26

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 27

Si il it f t t b i d i Similarity of states by crime dynamics

27

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 28

Si il it f ti i d b it ti Similarity of time periods by situations

28

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR

SLIDE 29

C l i Conclusions

Interactive and animated maps and graphs are not sufficient for analyzing large and

complex space-time data. Visual methods need to be augmented by computations.

With space-in-time and time-in-space SOMs we consider data from two different

perspectives: p p

places grouped by similar attribute dynamics
time intervals grouped by similar spatial distributions of attribute values
Colours of the groups reflect their similarity
Case studies demonstrate the value of the approach
for detecting the expected
for detecting the expected
for discovering the unexpected
Current work: non-tabular data, other clustering/projection methods

Live demo is possible

29

Gennady Andrienko 10 June 2010, presentation at EuroVis http://geoanalytics.net Bordeaux, FR