Authors Shahab Helmi shahab helmiucdenver edu Farnoush BanaeiKashani
Authors: • Shahab Helmi shahab. helmi@ucdenver. edu • Farnoush Banaei-Kashani farnoush. banaei-kashani@ucdenver. edu
2
Introduction Problem Definition Solution Experiments Related Work Conclusion & Future Work
Player 1 8 12 16 16 Player 2 13 15 16 14 Player 3 10 9 13 13 5
Player 1 8 12 16 16 Player 2 13 15 16 14 Player 1 Player 3 10 9 13 13 Player 3 8 12 16 13 13 6
Player 1 8 - 16 Player 1 8 12 16 Player 1 Player 3 12 16 13 13 Player 3 13 13 7
Player 1 8 12 16 15 16 12 16 Player 2 15 16 14 Player 3 9 13 13 Player 2 Player 1 8 14 8
9
10
Introduction Problem Definition Solution Experiments Related Work Conclusion & Future Work
Data preprocessing using the Spatial-Apriori property: Reducing the size of data Reducing the number of candidates Mining all frequent episodes using the MVS-FEM framework 12
If an episode is frequent, all of its sub-episodes are frequent If an episode is not frequent, none of its super-episodes are frequent Player 1 8 12 Player 1 Player 2 8 12 16 15 16 14 13
μ=3 14
Horizontal Growth Player 1 8 7 Vertical Growth Player 2 4 5 Player 1 8 Player 1 7 Player 2 4 Player 2 5 Player 1 8 Player 2 Player 1 Player 2 8 4 7 5 4 Player 1 8 7 Player 2 4 5 15
Apriori property Horizontally joinable episodes Player 1 8 7 9 3 Player 1 8 7 9 Player 1 7 9 3 16
At each iteration: 1. Generate all possible valid episodes by combining episodes generated in the Horizontal Growth step 2. Return frequent episodes There are too many combinations Not practical 17
Vertically joinable episodes Player 1 8 7 Player 2 4 5 Player 1 8 7 Player 3 1 6 Player 1 8 7 Player 2 4 5 Player 3 1 6 Player 2 4 Player 3 Player 4 Player 1 8 7 Player 2 4 5 5 Player 3 1 6 Player 4 9 9 18
Player 1 8 7 Player 2 4 5 Player 3 1 6 Player 1 8 7 9 Player 1 3 8 7 Player 2 4 5 4 Player 1 8 7 9 Player 2 4 5 4 Player 1 3 8 7 Player 2 4 5 4 19
Introduction Problem Definition Solution Experiments Related Work Conclusion & Future Work
Raw Processed GPS positions for 16 players (8 players per team) GPS positions for 14 players (7 players per team) Captured at 200 Hz Captured at 1 Hz 90 minutes 9 attributes: 3 attributes: Player # Timestamp Position of the player … 21
22
Parameters Symbol Default Value 7 Number of events per event sequence 2000 (seconds) Number of grid cells 256 Maximum time-span 8 seconds Minimum support 12 Performance evaluation measure: execution time 23
Symbol Default Value 7 2000 256 8 seconds 24
Symbol Default Value 2000 256 8 seconds 12 25
Symbol Default Value 7 2000 256 12 26
Introduction Problem Definition Solution Experiments Related Work Conclusion & Future Work
Trajectory query processing Frequent pattern mining Sport analytics 28
Models for trajectory pattern mining Activity recognition from a single trajectory Guessing the transportation mode multiple trajectories “Chasing" behavior 29
A B C A A, B A, C C, D C, F Transactional Data: Frequent itemset mining (order does not matter) Frequent sequence mining (order matters) Non-transactional data: Frequent episode mining from simple event sequences complex event sequences 30
Sport Analytics Predicting the next move in basketball Predicating the location of next shot in tennis Assessing the team formation in soccer Player analysis 31
Introduction Problem Definition Solution Experiments Related Work Conclusion & Future Work
Conclusion Introduced the co-movement pattern Introduced the “Spatial Apriori” property Proposed a preprocessing technique based on the Spatial Apriori property Introduced the MVS-FEM framework Proposed 3 algorithms for the MVS-FEM problem Future Work Efficient multi-resolution co-movement pattern mining Adding more spatial constraints to our framework Online mining algorithm 33
Q&A 34
Player 1 16 16 Player 2 7 7 Player 1 16 16 Player 1 Player 2 7 7 Player 2 Player 1 16 16 Player 2 7 7 16 16 7 7 35
Implementation: C# SQL Workstation: Intel Core-i 7 3. 6 GHz CPU 16 GB of memory running Windows 10 36
Symbol Default Value 7 2000 8 seconds 12 37
Symbol Default Value 7 256 8 seconds 12 38
- Slides: 38