Behavioral Knowledge from GPS Traces for Categorizing Mobile Users - - PowerPoint PPT Presentation

Modeling of Human Movement Behavioral Knowledge from GPS Traces for Categorizing Mobile Users

S O U R S E : W W W 2 0 1 7 A D V I S O R : J I A - L I N G KO H S P E A K E R : H S I U - Y I , C H U D AT E : 2 0 1 7 / 1 1 / 7

Outline

Introduction Method Experiment Conclusion

Introduction

Question

Can we map the knowledge of one known region to another unknown(target) region and use this knowledge to categorize the users in the target region?

Introduction

Goal

Labeled GPS log Unlabeled GPS log

User category label

Knowledge

Outline

Introduction Method Experiment Conclusion

Method

Method

User Trajectory Segment

<S[], W[], Traj_Win[]>

S[]: list of stay points, s = <lat, lon, Geotagg> W[]: list of waiting points, w = <lat,lon> Traj_Win[]: {S1, (x1,x2), (x2,y2), … ,S2}

Method

User Trajectory Segment

Method

Semantic Stay Point Taxonomy(SSPTaxonomy)

SSPTaxonomy:<N, Nc, W>

N: place type of the Taxonomy

Nc: associated code of the node place W: aggregated footprints of user

Method

Semantic Stay Point Taxonomy(SSPTaxonomy)

Method

User-Trace Summary(UTS)

NB = <G,Θ>,G=<V,E>

v1i: (Nc,ti) Nc: associated code of the node place ti: temporal value of the node ei: dependences between the vertices

Method

User-Trace Summary(UTS)

Bayesian Network

Method

Θx5|Pax5 = 0.46

Method

Θx4|Pax4*Θx5|Pax5*Θx2|Pax2 = 0.6*0.46*0.98=0.27048

Method

Temporal Common Sub-sequence, (TempCS)clustering algorithm

Similarity measure(Bhattacharyya distance):

DB(X4, X5)=-ln{[X4(0)X5(0)]1/2+[X4(1)X5(1)]1/2}=

• ln{[0.4*(0.4*0.32+0.6*0.54)]1/2+[0.6*(0.4*0.68+0.6*0.46)]1/2}

Method

Temporal Common Sub-sequence, (TempCS)clustering algorithm

NB1:X4X3X5X1X2 NB2:X4X1X6X5X2 Common stay points(Lc):X4X5X2X1 Common Sub-sequence(Ls):X4X5X2

Method

Similarity between NB1and NB2: SimSequence(NB1,NB2)= 3/4[DB(X4, X5)+ DB(X5, X2)]

Method

User Categorization

Classification task:

PVu= {p1,p2,…,pi} i: user-category pi: probability of the user u in category i

Method

User Categorization

Feature

f1: visit in types of places f2: Speed of movement or transportation mode f3: User Movement

Method

User Categorization

Bayesian network

When independent Weighting each of feature

Method

Transfer Learning

Method

Transfer Learning

Extract the parent’s code cp of a node c. Node c has n sibling, append n+1 along with the parent’s code cpn+1. Check whether the same place-type in and assign the same code if present. Generate Get the common taxonomy

Outline

Introduction Method Experiment Conclusion

Experiment

Dataset

Experiment

Accuracy of User-Classification

Experiment

Outline

Introduction Method Experiment Conclusion

Conclusion

Address the user categorization problem from the GPS traces of the users. Propose a framework to model individual’s movement patterns. Transfer knowledge base from one city domain to another unknown city.