Techniques and Challenges for Trajectory Prediction Slides credit: - - PowerPoint PPT Presentation

▶
techniques and challenges for
SMART_READER_LITE
LIVE PREVIEW

Techniques and Challenges for Trajectory Prediction Slides credit: - - PowerPoint PPT Presentation

Feb 14, 2024 8 likes •380 views

Where Next? Data Mining Techniques and Challenges for Trajectory Prediction Slides credit: Layla Pournajaf o Navigational services. o Traffic management. o Location-based advertising. Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti.

slide-1
SLIDE 1

Where Next? Data Mining Techniques and Challenges for Trajectory Prediction

Slides credit: Layla Pournajaf

slide-2
SLIDE 2
  • Navigational services.
  • Traffic management.
  • Location-based advertising.

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-3
SLIDE 3
  • Destination prediction
  • Path prediction with known destination
  • Path prediction with unknown destination
  • Similar to predicting next N locations

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-4
SLIDE 4
slide-5
SLIDE 5

Raw Trajectories

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-6
SLIDE 6

Raw Trajectories Preprocessed Trajectories

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-7
SLIDE 7

Raw Trajectories Preprocessed Trajectories Prediction Model

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-8
SLIDE 8

Source: www.openstreetmap.org

slide-9
SLIDE 9

Real-world data include raw trajectories of continuous GPS coordinates which are noisy and inaccurate!

Source: www.openstreetmap.org

slide-10
SLIDE 10
slide-11
SLIDE 11

Raw Trajectories Preprocessed Trajectories

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-12
SLIDE 12
  • Discretizing Time
  • 30 seconds, one hour
  • Temporal Representation
  • Location-series
  • Fixed-interval time-location series
  • Variable-interval time-location

series

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-13
SLIDE 13
  • Discretizing Location
  • Grid-based (uniform vs

hierarchical)

  • Mining Frequent Regions
  • Clustering
  • Semantic-based

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-14
SLIDE 14

Source: Xue, Andy Yuan, et al. "Destination prediction by sub-trajectory synthesis and privacy protection against such prediction." ICDE 2013.

Map of Beijing with 30 × 30 grid overlay: Each cell ≈ 1.78km2

slide-15
SLIDE 15
  • Clustering
  • DBScan
  • Hierarchical Clustering
  • Semantic-based
  • Using points of interests

Source: Lei, Po-Ruey, et al. "Exploring spatial-temporal trajectory model for location prediction." MDM 2011.

slide-16
SLIDE 16

Raw Trajectories Preprocessed Trajectories Prediction Models

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-17
SLIDE 17
  • Personalized / Individual-based:
  • Utilize only the history of one object to predict its future locations
  • General:
  • Utilize the history of all objects to predict future locations
slide-18
SLIDE 18
  • Model-based (formulate the movement of moving objects using

mathematical models)

  • Markov Models
  • Hidden Markov Models (Zhou et. al., ACM SIGKDD 2013)
  • Recursive Motion Function (Y

. Tao et. al., ACM SIGMOD 2004)

  • Deep learning models
  • Pattern-based (exploit pattern mining algorithms for prediction)
  • Sequential Pattern Mining (G. Yavas et. al., DKE 2005)
  • Trajectory Pattern Mining
  • Hybrid
  • Recursive Motion Function + Sequential Pattern Mining (H. Jeung et. al.,

ICDE 2008)

slide-19
SLIDE 19

Source: Xue, Andy Yuan, et al. "Destination prediction by sub-trajectory synthesis and privacy protection against such prediction." ICDE 2013.

slide-20
SLIDE 20

2 𝑞45 = 3 1 𝑞56 = 3

Source: Xue, Andy Yuan, et al. "Destination prediction by sub-trajectory synthesis and privacy protection against such prediction." ICDE 2013.

slide-21
SLIDE 21

Source: Xue, Andy Yuan, et al. "Destination prediction by sub-trajectory synthesis and privacy protection against such prediction." ICDE 2013.

slide-22
SLIDE 22

Partial Trajectory: <𝑠1, 𝑢1> , <𝑠2, 𝑢2>, …., <𝑠𝑑, 𝑢𝑑>

Prediction:

  • Having a partial trajectory (discretized) including the current

region 𝑠𝑑, find the most probable region at time point 𝑢𝑑+1

arg max P(𝑆𝑑+1 = 𝑠𝑑+1| 𝑠1 , … 𝑠𝑑)

𝑠𝑑+1 <?, 𝑢𝑑+1>

slide-23
SLIDE 23

Embedding Higher-Order Chains

  • Each new state depends on fixed-length

window of preceding state values

  • We can represent this as a first-order model

via state augmentation:

(N2 augmented states)

slide-24
SLIDE 24

Semi-Lazy Hidden Markov Approach (SIGKDD ‘13)

  • Find similar trajectories from historical trajectories (reference
  • bjects)
  • Build a hidden Markov Model on the fly (vs. eager or lazy

approach)

  • Self-correcting continuous prediction (real time)
  • Refine prediction model
  • Adjust weights for reference objects
slide-25
SLIDE 25
  • Model-based (formulate the movement of moving objects using

mathematical models)

  • Markov Models
  • Hidden Markov Models (Zhou et. al., ACM SIGKDD 2013)
  • Recursive Motion Function (Y

. Tao et. al., ACM SIGMOD 2004)

  • Deep learning models
  • Pattern-based (exploit pattern mining algorithms for prediction)
  • Sequential Pattern Mining (G. Yavas et. al., DKE 2005)
  • Trajectory Pattern Mining (Monreale et al ACM SIGKDD 2009)
  • Hybrid
  • Recursive Motion Function + Sequential Pattern Mining (H. Jeung et. al.,

ICDE 2008)

slide-26
SLIDE 26

1. Preprocess raw trajectories and extract frequent sequential patterns (T-Pattern)

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-27
SLIDE 27

1. Preprocess raw trajectories and extract frequent sequential patterns (T-Pattern) 2. Build a Prefix Tree (T-Pattern Tree)

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-28
SLIDE 28

1. Preprocess raw trajectories and extract frequent sequential patterns (T-Pattern) 2. Build a Prefix Tree (T-Pattern Tree) 3. Predict Next Location

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-29
SLIDE 29

<𝑦1, 𝑧1, 𝑢1> , <𝑦2, 𝑧2, 𝑢2>, …., <𝑦𝑜, 𝑧𝑜, 𝑢𝑜>

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-30
SLIDE 30
  • Two points match if one falls within a spatial neighborhood N()
  • f the other
  • Two transition times match if their temporal difference is ≤ τ

<𝑦1, 𝑧1, 𝑢1> , <𝑦2, 𝑧2, 𝑢2>, …., <𝑦𝑜, 𝑧𝑜, 𝑢𝑜>

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-31
SLIDE 31
  • Two points match if one falls within a spatial neighborhood N()
  • f the other
  • Two transition times match if their temporal difference is ≤ τ

<𝑦1, 𝑧1, 𝑢1> , <𝑦2, 𝑧2, 𝑢2>, …., <𝑦𝑜, 𝑧𝑜, 𝑢𝑜>

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

  • Calculate support for each T-pattern
slide-32
SLIDE 32

Generating all association rules from each T-pattern and using them to build a classifier is too expensive.

R1 R2 R3 R4

T-Pattern Rules

R1 R2 R3 R4 R1 R2 R3 R4 R1 R2 R3 R4

α1 α2 α3

Source: A. Monreale, F. Pinelli, R. Trasarti, F. Giannotti. WhereNext: a Location Predictor on Trajectory Pattern Mining. KDD 2009

slide-33
SLIDE 33

To avoid the rules generation, the T-Pattern set is organized as a prefix tree.

For Each node v

  • Id identifies the node v
  • Region is a spatial component of the T-Pattern
  • Support is the support of the T-pattern

For Each edge j

[a,b] correspond to the time interval αn of the T-Pattern

slide-34
SLIDE 34

Three steps:

  • 1. Search for best match
  • 2. Candidate generation
  • 3. Make predictions

Best Match Prediction

slide-35
SLIDE 35

Three steps:

  • 1. Search for best match
  • 2. Candidate generation
  • 3. Make predictions

The Best Match is the path having:

 the maximum path score using the time and location matching and support  at least one admissible prediction.

slide-36
SLIDE 36
  • Prediction errors (distance and time)
  • Prediction accuracy (precision and recall)
  • Prediction rate
slide-37
SLIDE 37

  • H. Jeung, Q. Liu, H. T

. Shen, and X. Zhou. A hybrid prediction model for moving objects. In Data Engineering, 2008. ICDE 2008. IEEE 24th International Conference on, pages 70-79. IEEE, 2008.

  • J. Krumm and E. Horvitz. Predestination: Inferring destinations from partial trajectories. In Ubiquitous Computing, pages 243-260. Springer, 2006.

  • A. Monreale, F. Pinelli, R. Trasarti, and F. Giannotti. Wherenext: a location predictor on trajectory pattern mining. In Proceedings of the 15th ACM SIGKDD

international conference on Knowledge discovery and data mining, pages 637-646. ACM, 2009.

  • G. Yavas, D. Katsaros, O. Ulusoy, and Y

. Manolopoulos. A data mining approach for location prediction in mobile environments. Data & Knowledge Engineering, 54(2):121-146, 2005.

Y . Tao, C. Faloutsos, D. Papadias, and B. Liu. Prediction and indexing of moving objects with unknown motion patterns. In Proceedings of the 2004 ACM SIGMOD international conference on Management of data, pages 611-622. ACM, 2004.

  • A. Y

. Xue, R. Zhang, Y . Zheng, X. Xie, J. Huang, and Z. Xu. Destination prediction by sub-trajectory synthesis and privacy protection against such prediction. In Data Engineering (ICDE), 2013 IEEE 29th International Conference on, pages 254-265. IEEE, 2013.