SenMinCom: Pervasive Distributed Dynamic Sensor Data Mining for

Effective Commerce

Outline1. What is SenMinCom ?2. Past Works & Why SenMinCom ?3. How SenMinCom ?4. SenMinCom’s Contributions 5. Current Methods v/s SenMinCom

6. SenMinCom’s Simulations1. Shopping Model2. Mobile Device Usage Model

7. Conclusion8. References9. Acknowledgements

What is SenMinCom [24] ? Independent units

that receive and respond to signals

Unobtrusive Cheaply available

computer

Sensing

contd… Process of sorting

through heap of data and picking out relevant gems

Mostly on data that have not been previously discovered

Mining

contd… Mobile commerce or

U commerce is the ability to conduct commerce using a cellular device

U-commerce because of its Ubiquitous-ness

mobile Comme

rce

contd…

unobtrusive Sensing

live Minin

g

mobile Commerce

Past Works & Why SenMinCom [24] ?

Sensors restricted to defense, environmental tracking, etc.Cellular phones limited to entertainment, commerce, etc.

contd…

Environmental monitoring [1-7]

• Enormous potential over the traditional invasive methods

Smart environments [8]

• Self-organizing, adaptive systems

Mobile Commerce [9-10]

• Discusses frameworks, applications

contd…

GypSii [11-12]

• Location service management

Mobile social networking [12-13]

• Share photos, send invitations, etc.

contd…

Sensors

Cellular phones

contd…

NTT DoCoMo’s Wellness Phone [14-15]

• Keeps track of runs, heart pulse, mini body fat calculator

Tsunami warning system [16]

• Register an earthquake & relays warning messages to the people in affected area

HealthGear [17]

• Set of non-invasive physiological sensors that monitors a patient’s health

contd…

Schwan’s for route sale drivers [18]

• Wirelessly record sales, issue receipts and track inventory

contd…

sensors

cellular phones

minin

g

contd…Shopping Scenario

•No method to get the real time shopping pattern

•No way to know about shoppers’ preference for products

•No effective way to lure shoppers’ before they leave store

•No system that benefits retail and consumer

contd…Mobile Usage Scenario

• Companies rely on quarterly surveys

• No method to get real-time mobile usage

contd…

automate the task of market surveys

up-to-the-minute information

effective progress

How SenMinCom [24] ?

Centralized static data mining

Distributed dynamic data mining

contd…

Centralized static data mining

Off-line data mining

Time & resource intensive

Sink processes & analyzes

Sensor tracks & forwards

contd…

Mobile agent•Formed by code & data•Clone & migrate•Reduce network load•Overcome network latency

Data mining•Reveal patterns•Easy to perceive, interpret, manipulate •Mining accomplished on real time data rather than on a snapshot

Sensor •Achieves unobtrusive sensing•Gathers +Processes +Communicates

contd…

SenMinCom [24]

Sensed data

Mobile agent

Data mining

SenMinCom’s Contributions [24]

contd…

Centralized location

Regional server

Network

Store1…n

Sensor-ized area

Customer

contd…

Aggregator nodes

Sensor-ized area

contd…

Distributed Dynamic

Mining System (DDMS)

Reduces communication effort [19]

Dimensionality reduction of data mining

Mining on fresh data

Less resource intensive

contd…

A DDMS is a set of transactions <T, t> where 'T' is a purchase or product information event and 't' represents the time and date of the occurrence of T.

DDMSM = { <T1 ,t1>...<Tn, tn>} where M is the recorded Mac Id of the customer's cell phone.

System will have pre-defined rule base from which the distinction of customers is achieved.

contd…

i. Define Rules and corresponding parameters

for each Rule

ii. for shopper (Mac Id) m=1 to M

i. Identify DDMSm from set of DDMS

ii. for segment r=1 to Ri. Select Ruler from the set of Rules, R

ii. If(Ruler ⊆ DDMSm ) add shopper m to group r

iii. The R groups of shoppers are the segments

contd…If(thisNode = = firstAggregator)

MA migrates toward firstAggregatorElse if( (thisNode = = nextAggregator) &&

(nextAggregator != lastAggregator) )MA collects sensed raw data and does local miningSet nextAggregator in the MA packetMA migrates towards next aggregator

Else if(thisNode = = lastAggregator)MA collects sensed dataMA migrates back to sink

Current Methods v/s SenMinCom [24]

Communicating messages consume far more energy than processing it [21]

• Mining done at aggregator

Mined results don’t affect real-world situation [22]

• Mining takes place on fresh data

Link bandwidth of wireless sensor network very less [20]

• Agent carries only the result set

contd…

Central mining costly in terms of communication and storage [23]

• Task of mining distributed on all aggregators

Sensor nodes passive

• Sensor made an active device

SenMinCom’s Simulations

Shopping ModelMobile Device Usage

Model

Shopping Model [24]

Random shoppers have no strong intention to purchase something, and just wander among aisles a.k.a. window shoppers

Rational shoppers visiting a store, know clearly what they need a.k.a prompt shoppers

Recurrent or regular customers are customers who visit the store often. They can be further divided into Customers with higher purchasing power Customers with lower purchasing power

contd…

Example Book store company e.g. Barnes & Nobles Store modeled on SenMinCom architecture

Result Customers shopping & checkout patterns

dynamically tracked

contd…

Features

1. Aisle wise real time products distribution

2. Reveals aisle popularity

Consequences

3. Restacking products

4. Maximize selling

Ais

le I

Ais

le II

Ais

le II

I

Ais

le IV

Ais

le V

Ais

le V

I

Ais

le V

II

Ais

le V

III

Ais

le IX

Ais

le X

0

10

20

30

40

50

60

Ais

le I

Ais

le II

Ais

le II

I

Ais

le IV

Ais

le V

Ais

le V

I

Ais

le V

II

Ais

le V

III

Ais

le IX

Ais

le X

Aisles

Prod

ucts

Aisle

contd…

Features

1. Aisle wise real time products distribution at separate time intervals

2. Aisle popularity

Consequences

3. Restacking products according to different hours of a day, days in a week, etc.

Ais

le I

Ais

le II

Ais

le II

I

Ais

le IV

Ais

le V

Ais

le V

I

Ais

le V

II

Ais

le V

III A

isle

IX

Ais

le X

Ais

le I

Ais

le II

Ais

le II

I

Ais

le IV

Ais

le V

Ais

le V

I

Ais

le V

II

Ais

le V

III

Ais

le IX

0

10

20

30

40

50

60

70

80

Aisle I Aisle II Aisle III Aisle IV Aisle V Aisle VI AisleVII

AisleVIII

Aisle IX Aisle X

Aisles

Prod

ucts

Time 1

Time 2

contd…

Features

1. Reveals customers purchasing power

2. Categorize customers

Consequences

3. Directed products promotion

30

34 34

38

43

46

49 49 50

42

0

10

20

30

40

50

60

Ali

ce

Che

tan

Chi

rayu

Sarj

i

Sara

h

Mar

k

Mon

ica

Muk

und

Sund

ar

Mou

nica

Customers

Prod

ucts

Shopping Share

contd…

Feature

1. Products lifted to checked out

Consequences

2. With shopping history product promotion offers

3. Customer

Ali

ce

Che

tan

Chi

rayu

Mar

k

Mon

ica M

ouni

ca

Muk

und

Sara

h Sund

ar

Ali

ce Che

tan

Chi

rayu

Mar

k Mon

ica

Mou

nica

Muk

und

Sara

h

Sarj

i

Sund

ar

Sarj

i0

10

20

30

40

50

60

Ali

ce

Che

tan

Chi

rayu

Mar

k

Mon

ica

Mou

nica

Muk

und

Sara

h

Sarj

i

Sund

ar

Customers

Prod

ucts

Checkout Share

Shopping Share

contd…

Feature

1. Products lifted to checked out customer level

Consequences

2. Shopping history leads product promotion offers

3. Products picked to checked out share

4. Aisle movement pattern

3

3

3

5

4

4

11

1

4

2

0

3

5

4

5

3

10

1

4

0

0 1 2 3 4 5 6 7 8 9 10 11 12

Aisle I

Aisle II

Aisle III

Aisle IV

Aisle V

Aisle VI

Aisle VII

Aisle VIII

Aisle IX

Aisle X

Prod

ucts

Aisles

Checkout

Shopping

Mobile Device Usage Model Popular cellular phone cravings

Brand popularity where the people are attracted or loyal towards a company

For a cell phone company, popularity of a given model or total volume of their models

Cellular phone usage among an age group Educational period is a stage among the age

group of 18-28, generally students attending schools, colleges, and universities.

Working period, among the age group 28-60

contd…

Example Georgia State University Campus Area modeled on SenMinCom architecture

Result Students real time device usage scenario Manual device survey avoided

contd…M

otor

ola

LG

Sam

sung

Nok

ia App

le

Bla

ckB

erry

0

10

20

Mot

orol

a

LG

Sam

sung

Nok

ia

App

le

Bla

ckB

erry

Mobile Device

Mobile Devices GSU Plaza

contd…M

otor

ola LG

Sam

sung Nok

ia

App

le

Bla

ckB

erry

0

10

20

30

Mot

orol

a

LG

Sam

sung

Nok

ia

App

le

Bla

ckB

erry

Mobile Device

Mobile Devices GSU Student Center

contd…M

otor

ola

LG

Sam

sung

Nok

ia

App

le

Bla

ckB

erry

0

10

20

30

40

Mot

orol

a

LG

Sam

sung

Nok

ia

App

le

Bla

ckB

erry

Mobile Device

Popular Mobile Devices @ GSU

contd…

Features

1. Area wide popular mobile models

2. Total mobile device usage scenario

Consequences

3. Real time mobile popularity

4. Brand consideration leads to streamlining promotions

contd…

Feature

1. Various mobile models of a brand

Consequences

2. Popularity of models

3. Reasons like cost, intriguing features, etc. revealed

0

7

0 0

9

6

0

10L

G S

hine

LG

Ax3

90

LG

Sco

op

LG

env

LG

Vx8

500

LG

Vx8

550

LG Models

Vol

ume

LG

contd…3

0

5

4

1

0

0

10

Mot

o R

azr V

3m

Mot

o R

azr V

9

Mot

o K

rzr k

1m

Mot

o R

azr v

8

Mot

o Q

9c

Mot

o V

365

Motorola Models

Vol

ume

Motorola

Mot

orol

a

LG

Sam

sung

Nok

ia

App

le

Bla

ckB

erry

0

10

20

30

40

Mot

orol

a

LG

Sam

sung

Nok

ia

App

le

Bla

ckB

erry

Mobile Device

Motorola Volume Usage Popular Mobile Devices

contd…

Feature

1. Market share of cell phone models

Consequences

2. Timeline based share of model

3. Provide insight for a newly released model

0

2

8

19

17

11

0

10

20

Tim

e I

Tim

e II

Tim

e II

I

Apple Models

Vol

ume

iPhone 3

iPhone 2

contd…

Feature

1. Mobile usage of new cell phone models

Consequences

2. Crosscheck their marketing campaign

3. Peoples’ current mobile preferences

2

8

1

3

4

0

10M

oto

Kra

zrK

1m

iPho

ne 3

G

LG

Vx8

500

LG

Vx8

550

LG

Ax3

90New Sightings

Vol

ume

New

Conclusion

SenMinCom [24]

Sensors extended to retail

Real time pervasive system

Data centric

Real time analysis of business

References1. Mainwaring, A., Polastre, J., Szewczyk, R., Culler, D., Anderson, J.,“Wireless

Sensor Networks for Habitat Monitoring”, Proceedings of the 1st ACM International Workshop on Wireless Sensor Networks and Applications, 2002, pp.88-97.

2. Warrior, J., “Smart Sensor Networks of the Future”, Sensors Magazine, March 1997.

3. Pottie, G.J., Kaiser, W.J., “Wireless Integrated Network Sensors”, Communications of the ACM, vol. 43, no. 5, pp.551-55 8, May 2000.

4. Cerpa, A., Elson, J., Estrin, D., Girod, L., Hamilton M., Zhao, J., “Habitat monitoring: Application driver for wireless communications technology”, 2001 ACM SIGCOMM Workshop on data Communications in Latin America and the Caribbean, Costa Rica, April 2001.

5. Werner-Allen, G., Johnson, J., Ruiz, M., Lees, J., Welsh, M., “Monitoring volcanic eruptions with a wireless sensor network”, Wireless Sensor Networks, 2005. Proceedings of the second European Workshop, 2005, pp.108-120.

6. Intel Research Sensor Network Operation, http://intel.com/research/exploratory/wireless_sensors.htm .

7. Shih, E., Cho, S., Ickes, N., Min, R., Sinha, A., Wang, A., Chandrakasan, A., “Physical layer driven protocol and algorithm design for energy-efficient wireless sensor networks”, Proceedings of ACM MobiCom’01, Rome, Italy, July 2001, pp.271-286.

contd…8. Herring, C., Kaplan, S., “Component-based software systems for smart

environments, IEEE Personal Communications, October 2000, pp. 60-61.

9. Varshney, U., Vetter, R., “Framework, Applications, and Networking Support for M-commerce”, ACM/Kluwer Journal on Mobile Network and Applications (MONET), June 2002.

10. Varshney, U., Vetter, R., Kalakota, R.,”Mobile Commerce: A New Frontier”, IEEE Computer, 2000, 22(10), pp.32-38.

11. GyPSii Webtop, http://www.gypsii.com/

12. Social Networking moves to the cell phone, http://www.nytimes.com/2008/03/06/technology/06wireless.html?_r=1&oref=slogin

13. Social Network Zingku, http://www.infoworld.com/article/07/09/28/Google-buys-Zingku-mobile-social-networking-service_1.html

14. NTT DoCoMo Newsletter, Mobility, Adding the Human Touch to Communication, http://www.nttdocomo.com/binary/about/mobility_doc_15.pdf

15. New Cell phone doubles as personal trainer and shrink, http://tech.yahoo.com/blogs/null/50133

contd…16. MedDay has Breakthrough solution for Tsunami Warning System

based on disease detection and management system, http://www.hoise.com/vmw/05/articles/vmw/LV-VM-02-05-16.html

17. Oliver, N., Flores-Mangas, F., HealthGear: A Real-time Wearable System for Monitoring and Analyzing Physiological Signals, Proceedings of the International Workshop on Wearable and Implantable Body Sensor Networks (BSN ’06), 2006

18. Schwan’s, http://www.bluetooth.com/NR/rdonlyres/826F390E-C82E-43A4-9810-B1D7291A275D/0/schwans.pdf

19. Goel, S., and Imielinski, T., “Prediction-based monitoring in sensor networks: Taking lessons from mpeg”, ACM Computer Communication Review, 31(5), 2001.

20. Chen M., Kwon, T., Choi, Y., “Data Dissemination based on Mobile Agent in Wireless Sensor Networks”, Proceedings of the IEEE Conference on Local Computer Networks 30th anniversary (LCN '05).

21. Chen M., Kwon, T., Yuan, Y., Leung V.C.M., “Mobile Agent Based Wireless Sensor Networks”, Journal of Computers, Vol 1, No. 1., April 2006

contd…22. Ong, K., Zhang, Z., Ng, W., Lim, E., “Agents and Stream Data

Mining: A New Perspective”, IEEE Intelligent Systems, June 2005.

23. Bontempi, G., Borgne, Y., “An Adaptive Modular Approach to the mining of Sensor Network Data”, 2005 SIAM International Conference on Data Mining, April 2005.

24. Hiremath, N., Zhang, Y., “SenMinCom: Pervasive Distributed Dynamic Sensor Data Mining for Effective Commerce,” Proceedings of 2008 IEEE International Conference on Granular Computing (GrC 2008), Aug 2008