CellTrans: Private Car or Public Transportation? Infer Users Main - - PowerPoint PPT Presentation

▶
celltrans private car or public transportation
SMART_READER_LITE
LIVE PREVIEW

CellTrans: Private Car or Public Transportation? Infer Users Main - - PowerPoint PPT Presentation

Sep 20, 2022 300 likes •671 views

CellTrans: Private Car or Public Transportation? Infer Users Main Transportation Modes at Urban Scale with Cellular Data Yi Zhao*, Xu Wang*, Jianbo Li , Desheng Zhang , Zheng Yang* *Tsinghua University, Qingdao University,

slide-1
SLIDE 1

CellTrans: Private Car or Public Transportation? Infer Users’ Main Transportation Modes at Urban Scale with Cellular Data

Yi Zhao*, Xu Wang*, Jianbo Li†, Desheng Zhang ‡ , Zheng Yang*

*Tsinghua University, †Qingdao University, ‡Rutgers University

slide-2
SLIDE 2

Motivation

Understanding citizens’ main transportation modes at urban scale is beneficial to a range of applications.

2

City Planning Transportation Management LBS

slide-3
SLIDE 3

Motivation

The inference of trajectory’s transportation modes has been well-studied

  • n GPS and phone sensor data, which are collected in a limited scale.

3

GPS Data Geolife dataset[1]: 182 users Sensor Data SHL dataset[2]: 3 users

[1] Yu Zheng, Yukun Chen, Quannan Li, Xing Xie, and Wei-Ying Ma. 2010. Understanding Transportation Modes Based on GPS Data for Web Applications. ACM Trans. Web 4, 1, Article 1 (Jan. 2010), [2] Lin Wang, Hristijan Gjoreskia, Kazuya Murao, Tsuyoshi Okita, and Daniel Roggen. 2018. Summary of the Sussex-Huawei Locomotion-Transportation Recognition Challenge. In Proceedings of UbiComp 2018.

slide-4
SLIDE 4

Cellular networks

Fast development of cellular networks:

  • Large scale, both spatially and temporally.
  • Low cost, already collected for billing purposes.

8,918,157,500

Mobile Devices

7,687,783,109 5,123,988,900

World Population Unique Subscribers

4

slide-5
SLIDE 5

Question

Can cellular data be used to infer users’ main transportation modes?

  • Direct solution based on previous methods:

Find Trips Infer Mode Main Mode

However, this direct solution does not work.

5

slide-6
SLIDE 6

The direct solution does not work for cellular data:

Coarse spatial granularity Irregular temporal sampling

6

slide-7
SLIDE 7

CellTrans

Stay Stay Stay Trip Trip

  • Instead of focusing on each trip,

CellTrans considers a long period

  • f users’ location records.
  • The expansion of observation time

can compensate for the coarse spatiotemporal granularity of cellular data.

7

slide-8
SLIDE 8

Framework of CellTrans

8

slide-9
SLIDE 9

Dataset

9

We base our design on two large-scale cellular datasets from different cities: Shenyang and Dalian.

Shenyang Dalian

slide-10
SLIDE 10

Trajectory Processing

Parsing users’ raw cellular data into stays and trips.[1]

[1] S. Jiang, J. Ferreira, and M. C. Gonzalez. 2017. Activity-Based Human Mobility Patterns Inferred from Mobile Phone Data: A Case Study of Singapore. IEEE Transactions on Big Data 3, 2 (June 2017), 208–219

Stay Stay Stay Trip Trip

  • Stays usually correspond to users’

activities like resting at home or working at office.

  • Trips are trajectory segments

when users travel from one stay region to another by some transportation means

10

slide-11
SLIDE 11

Mobility Features

11

Stays & Trips Mobility Features Main Mode

Extract Infer Movement Range Trip Statistics User Behavior

slide-12
SLIDE 12

Mobility Features: Movement Range

12

It is easier for people driving car to visit more and further places compared to people taking public transportation.

  • 1. Radius of Gyration

rg=rg rg=rg ncluster=4

  • 2. # of Stay Clusters

ncluster=2

  • 3. Convex Hull Area

a=0.5*rg*rg a=0

slide-13
SLIDE 13

Mobility Features: Trip Statistics

13

The high-level statistics of trips can provide useful information to infer users’ main transportation modes.

  • 4. # of Trips
  • 5. # of Night Trips
  • 6. Average Speed
slide-14
SLIDE 14

Mobility Features: User Behavior

14

The living pattern and economical status may be different between users

  • f different modes.
  • 7. Network Access during Trip
  • 8. Schedule
  • 9. House Price
slide-15
SLIDE 15

Mode Inference Model

Scenario 1: With Labeled Users. We assume that partial users’ actual modes are known, so a supervised model can be trained.

15

  • Radius of Gyration
  • # of Stay Clusters
  • Area of Convex Hull

…

Mobility Features

  • SVM
  • Random Forest
  • MLP

…

Supervised Models

slide-16
SLIDE 16

Mode Inference Model

Scenario 2: Without Labeled Users:

16

Car or Public trans. users

…

Clustering

… … …

slide-17
SLIDE 17

Model Training

Mode Inference Model

Scenario 2: Without Labeled Users:

17

Labeled Users

…

SVM RF MLP …

…

City A

…

City B

… …

slide-18
SLIDE 18

Evaluation

Groundtruth:

18

  • ws/mapapi/navigation/auto
  • ws/transfer/navigation/auto
  • …
  • ws/mapapi/navigation/bus/ext
  • ws/mapapi/realtimebus/linestation
  • …

Shenyang Dalian

slide-19
SLIDE 19

Evaluation: Scenario 1

19

MFR Data Mobility Features SVM Main Mode Trips Previous Methods Aggregate

slide-20
SLIDE 20

Evaluation: Scenario 1

20

Shenyang Dalian

  • In Shenyang, CellTrans improves the accuracy by 20%.
  • In Dalian, CellTrans improves the accuracy by 19%.
slide-21
SLIDE 21

Labeled Users

  • Evaluate the trained model at urban scale.

Evaluation: Scenario 1

21

… …

Evaluate on All Users Model Training SVM Model

slide-22
SLIDE 22

Evaluation: Scenario 1

22

Distribution of car/public transportation users’ homes:

  • A: High-end residential areas -> More car users.
  • B: Universities -> More public transportation users.

Shenyang, car users Shenyang, public transportation users

slide-23
SLIDE 23

Evaluation: Scenario 2

23

Shenyang Dalian

  • Our methods outperform previous methods in both cities.
  • The transferred model achieves the best results.
slide-24
SLIDE 24

How important is each feature? -> The coefficients in Linear SVM.

  • Some features are important in both cities.
  • Some features are important in one city.

Evaluation: Feature Importance

24

Dalian Shenyang

slide-25
SLIDE 25

Evaluation: Feature Distribution

25

  • Some features have obviously different distribution between two

modes.

slide-26
SLIDE 26

Evaluation: Feature Distribution

26

  • Some features have similar distribution, but they are still helpful to

differentiate main transportation modes.

slide-27
SLIDE 27

Summary

  • We present CellTrans, a novel framework to survey users’ main

transportation modes (public transportation or private car) at urban scale.

  • We devise techniques to extract various mobility features from noisy

cellular data that are pertinent to users’ transportation modes.

  • We carry out comprehensive experiments to evaluate the performance
  • f CellTrans on two large-scale cellular datasets.
slide-28
SLIDE 28
slide-29
SLIDE 29

Dataset

1 2 3

a

29

slide-30
SLIDE 30

Dataset

30

The distribution of cellular data is uneven.

slide-31
SLIDE 31

Preprocessing

The preprocessing module deals with two problems of cellular data:

31

Oscillation[1] Bursty Sampling[2]

[1] Ling Qi, Yuanyuan Qiao, Fehmi Ben Abdesslem, Zhanyu Ma, and Jie Yang. 2016. Oscillation Resolution for Massive Cell Phone Traffic Data. MobiData ’16 [2] Yi Zhao, Zimu Zhou, Xu Wang, Tongtong Liu, Yunhao Liu, and Zheng Yang. 2019. CellTradeMap: Delineating Trade Areas for Urban Commercial Districts with Cellular

  • Networks. INFOCOM 2019.
slide-32
SLIDE 32

Mode Inference Model

Scenario 1: With Labeled Users:

32

slide-33
SLIDE 33

Value of Rg

33

CDF of rg for all users and car/pub. users.

Shenyang Dalian

slide-34
SLIDE 34

Selection of k in K-means

34

Accuracy with k.

Shenyang Dalian

slide-35
SLIDE 35

How many labeled users do we need?

35

Shenyang Dalian