Semantic Trajectories

1
Semantic Trajectories
GeoPKDD

• Stefano Spaccapietra
• SeCoGIS 2009, Gramado

2
Trajectories Multiple Common Senses

• Mathematical Trajectories
• Metaphorical Trajectories
• Geographical Trajectories
• Spatio-temporal Trajectories

3
Mathematical Trajectories

• A predictable path of a moving object

4
Metaphorical Trajectories

• An evolutionary path in some abstract space
• e.g. a 3D professional career space
• ltposition,
  institution, timegt
• 
  with stepwise variability

Institution
5
Naive Geographical Trajectories

• A travel in "geographical" space, i.e. a discrete
  space occupied by spatial objects
• gt time-varying attribute "current city"
• a special case of metaphorical trajectories

stop
6
Spatio-Temporal Trajectories

• Travel in physical space, i.e. a continuous space
  where position is defined using spatial
  coordinates
• Physical Movement
• Birds tracking

7
Let us focus onSpatio-Temporal Trajectories
8
Abundance of ST movement data

• GPS devices, sensors and alike nowadays allow
  capturing the position of moving objects.
• Movement can thus be recorded, either
  continuously or discretely, as a novel
  spatio-temporal feature of the moving objects.
• A large number of applications in a variety of
  domains are interested in analyzing movement of
  some type of objects or phenomena.
• city traffic management and planning
• goods delivery
• social habits of populations
• epidemic monitoring, pollution monitoring
• animal tracking
• ..

9
Trajectories A semantic view of movement

• Movement (continuous) F (t) ? space
• . usually, you don't "keep moving", you
  go from one place to another place
• Semantic units of movement (discrete)
  movements with a purpose trajectories

• From home to university
• From university back home
• From class room to cafeteria

10
The European GeoPKDD project (2005-2009)

• GeoPKDD Geographic Privacy-aware Knowledge
  Discovery and Delivery (http//geopkdd.isti.cnr.it
  /)
• Goal to develop theory, techniques and systems
  for geographic knowledge discovery, based on new
  privacy-preserving methods for extracting
  knowledge from large amounts of raw data
  referenced in space and time.
• Specifically, to devise data warehousing and data
  mining methods for trajectories of moving
  objects such methods are designed to preserve
  the privacy of the source sensitive data.

11
The GeoPKDD Scenario
Traffic
Management
Management
Accessibility of
Accessibility of
services
services
Mobility
Mobility
evolution
evolution
Urban planning
Urban planning
.

.

Telecommunication
company
Public administration or
business companies
Privacy
aware
-
Data mining
GeoKnowledge
GeoKnowledge
Interpretation and visualization of patterns
using geography and domain knowledge,
ST Patterns
warehouse
warehouse
Trajectory Warehouse
Trajectories warehouse
Trajectories warehouse
Privacy enforcement
12
Types of Queries in GeoPKDD

• Database queries
• How many cars are currently traveling along the
  Champs-Elysées avenue?
• Data Mining queries
• Which are the heaviest congestion areas in the
  city on weekdays? (e.g. use of clustering)
• Which are the sequences of places most visited on
  Sunday mornings? (use of patterns)
• Analysis/Reasoning queries
• Which are the suspicious/dangerous movements of
  visitors in a given recreational area?

13
MODAP (2009-2012)

• Mobility, Data mining And Privacy
• An EU coordinated action focus on dissemination
• www.modap.org
• Privacy risks associated with the mobility
  behavior of people are still unclear, and it is
  not possible for mobility data mining technology
  to thrive without sound privacy measures and
  standards for data collection, and data/knowledge
  publishing.
• MODAP aims to continue the efforts of GeoPKDD by
  coordinating and boosting the research activities
  in the intersection of mobility, data mining, and
  privacy.
• MODAP welcomes new members (active members and
  observers).

14
Trajectory Modeling

• A trajectory is a spatio-temporal object rather
  than a spatio-temporal property
• A spatio-temporal object with some generic
  features and some semantic features
• generic application independent
• semantic application dependent

15
Basic Definition

• (Point-based) Trajectory
• the user-defined record of the evolution of the
  position (perceived as a point) of an object
  traveling in space during a given time interval
  in order to achieve a given goal.
• trajectory F tbegin, tend ?
  space
• A trajectory is a semantic object, different
  from the corresponding physical object built on
  raw data
• Raw data the physical positioning acquired
  using GPS as a sequence of (point, instant) pairs
  (sample points)
• Raw movement data frequently needs to be cleaned
  before it can be used

16
From (x,y,t) Movement to Trajectories
17
Interpretations of Trajectories
(EPFL Metro Station, 840)?(INM202,
850)(INM202,1030)?(INM0,1032)(INM0,1058)?(IN
M202, 1100)(INM202,1200)?(Parmentier,1210)
denotational
(EPFL Metro Station, 840)?(seminar room,
850)(seminar room,1030)?(cafeteria,1032)(cafe
teria,1058)?(seminar room, 1100)(seminar
room,1200)?(restaurant,1210)
functional
18
Queries

• On movement data
• When cars stopped today at position (x,y)?
• Which cars stopped today at position (x,y)?
• On semantic trajectory data
• Which cars stopped today at at a gas station?
• For a given petrol company, return the number of
  cars that stopped today at a gas station owned by
  this companys retailers

19
Trajectory Characterization

• Attributes
• e.g. the goal of the trajectory (e.g. visit a
  customer)
• Links to other objects
• e.g. to the customer visited with this trajectory
• Constraints
• e.g. the trajectory of a car is constrained by
  the road network
• Begin End Points
• Delimit a trajectory
• Spatial type Point
• Temporal type Instant
• Topological inside links to spatial objects
• e.g. inside a City
• Attributes Links to other objects
  Constraints

20
Trajectory Components Stops

• Stop(s)
• Point
• Time interval
• Topological inside (or equal) link to a spatial
  object
• e.g. inside a City
• Attributes?
• Links to other objects?
• e.g. a RentalCarCompany, several Customers
• Constraints?

21
Trajectory Components Moves

• Move(s)
• Time varying point
• Time interval
• Topological inside (or equal) link(s) to (a)
  spatial object(s)
• e.g. the move follows part of Highway A3
• Attributes
• Non-varying attributes, i.e. attributes that have
  a fixed value during the whole duration of the
  move (e.g. duration)
• Varying attributes, i.e. attributes whose value
  varies during the move (e.g. the altitude of the
  plane)
• Links to other objects?
• e.g. the move was done with other persons
• Fixed link, i.e. the link links the same unique
  object during the whole move, e.g. link to the
  car used during the move
• Varying link, e.g. link to the transport means
  used during the move attached to object instance
  walking for the first 10mn, then attached to
  instance bus for the next 15mn, then ...
• Constraints

22
More Basic Definitions

• Stop
• a part of a trajectory defined by the
  user/application to be a stop, assuming the
  following constraints are satisfied
• during a stop, traveling is suspended (the
  traveling object does not move wrt the goal of
  achieving its travel) the spatial range of a
  stop is a single point
• the stop has some duration (its temporal extent
  is a non-empty time interval) the temporal
  extents of two stops are disjoint
• NB conceptual stops are different from physical
  stops
• Move
• a part of a trajectory between two consecutive
  stops, or between the starting point (begin)
  and the first stop, or between the last stop
  and the end point.
• the temporal extent of a move is a non-empty
  time interval
• the spatial extent of a move is a
  spatio-temporal line (not a point)
• Begin, End
• the two extremities of a trajectory (point,
  instant)

23
Stops and Moves Semantic Trajectories
A day in Paris
Airport Ibis Hotel
Eiffel Tower Louvre 0800-0830
0900-1200 1300-1500 1600-1800


24
Trajectory Components Episodes

• Generalizes stops and moves
• Time varying point
• Time interval
• Suitable for e.g. animal trajectories
• Episodes defined by activity
• sleeping
• searching for food
• reacting to an alert (escaping)

25
TrajectoryReconstruction
26
From raw data to semantic trajectories
27
Raw Data Cleaning
Input
Output
Methods filtering, smoothing, outliers removal,
missing points interpolation map-matching,
data compression, etc.
28
Trajectory Identification
Input Cleaned raw data
Output Trajectory segments (Begin, End)
Methods various segmentation algorithms (based
on spatial gaps, temporal gaps, time intervals,
time series, )
29
Trajectory Structure
Input Trajectories
Output Trajectory sub-segments (Stop, Move)
Methods various stop identification algorithms
(based on velocity, density, )
30
Velocity-based stop identification
31
Determining Stops and Moves

• User-defined
• Geometric computation
• Stops are abstractions (e.g. centroid) of an
  area where the moving object/point stays for a
  certain period of time
• Geometric Semantic computation
• Stops are all points representing selected
  objects of a certain type (hotel, restaurant, )
  where the moving object stays for a period whose
  duration is above a certain threshold
• Relevant objects may be defined at the type
  level (e.g. hotel, restaurant, ) or at the
  instance level (selected locations, e.g. customer
  premises)
• .

32
Semantic Enrichment
Input Structured Trajectories
Output Semantic trajectories
Methods use relationships of each structured
component (begin, end, stops, moves, ) to
application knowledge, i.e. meaningful objects
32
33
Capturing Trajectory Semantics
34
The design pattern
Its personalization
The hooks
35
A schema for bird monitoring
36
A Traffic Database with Trajectories
37
Trajectory Mining
38
Combining Space, Time and Semantics
SpatioTemporal (Flock) Pattern
Trajectory Semantic (Flock) Pattern
R
H
Touristic Place
Hotel
TP
Restaurant
Semantic trajectory mining pattern Hotel ?
TouristicPlace ? cross(A)
39
Convergence Patterns
SpatioTemporal Pattern
Semantic Pattern
T4
S
T1
S
T2
S
C
C
T3
T5
S
S
School
S
Semantic trajectory mining pattern School to C
40
Example Association Rules on Stops

• SELECT associateStop (minsup0.05, minconf0.4,
• timeGweekdayWeekend, stopGinstance)
• FROM stopTable
• Patterns
• PlazaHotelweekday ? Montmartreweekday
  (s0.08)
  (c0.47)
• Trajectories stopping at the PlazaHotel also stop
  at Montmartre (in any order)

41
Example Sequential Patterns on Moves

• SELECT sequentialMove (minsup0.03,
  timeG0800-1200, 1201-1800, 1801-2300,
  stopGinstance)
• FROM moveTable
• Patterns
• PlazaHotel - EiffelTower0800-1200 ?
  Louvre - PlazaHotel1201-1800 (s0.06 )
• CentralHotel - NotreDame0800-1200,
  Invalides -EiffelTower1201-1800
• ? EiffelTower-CentralHotel1801-2300
  (s0.04)
• In this order

42
Moving Object Behavior

• From semantic trajectories we can aim at
  understanding the behavior of moving objects
• Example converging patterns of people may
  indicate an intention to perform a joint action
• Ethically appropriate or not?
• Example from trajectories or firemen we may
  guess how a fire situation evolves
• Ethically appropriate or not?
• ? Privacy preserving analysis methods
• Join us at www.modap.org

43
Thanks