measuresandmetrics_final - T race Analysis Clustering and...

Trace Analysis, Clustering and
Network Theory
NOMADS

Usage Trace in RAW Form

2007 11 17 00:00:03 EST b430bar-win-ap1200-1 [info] 189341:
Nov 17 00:00:02 EST: %DOT11-6- ASSOC : Interface
Dot11Radio0, Station 001b.fcb2.4fc9 Reassociated
KEY_MGMT[NONE]

2007 11 17 00:00:19 EST elm-authgw-bs2100-1 [notice]
user_tracking:
event=user_login_successful&loglevel=notice&obj=user&ipadd
r=10.249.52.174&name= bob &msg=Login RADIUS user dung
on Primary RADIUS server at [ 00:1b:fc:b2:4f:c9 ]/10.249.52.174
as role Authenticated, login time = 2007-11-17 00:00:19,
sessionID = 00:0E:0C:33:08:BA:119527561961645&

Information Extraction

HOST_MAC VARCHAR2(20)

START_TIME DATE

END_TIME DATE

AAP_NAME VARCHAR2(150)

START_TIMESTAMP NUMBER(38)

END_TIMESTAMP NUMBER(38)

ATIMEZONE VARCHAR2(20)

RECORD_TYPE NUMBER(38)

ATRANSACTION_ID NUMBER(38)

AAP_BLDG VARCHAR2(20)

DAP_BLDG VARCHAR2(20)

DAP_NAME VARCHAR2(150)

DTIMEZONE VARCHAR2(20)

DTRANSACTION_ID NUMBER(38)

ROAM_MAC VARCHAR2(20)

Udayan Mapping with UF Phone Directory

Number Theory and Counting

My Experience  Logic Comes First

Focus on the Idea and NOT on the Data

Selection of right data

Selection of right Format

Data Representation should be correct

Start with a small Sample (training set) and move

Most Common Form
Start Time Location Duration
8175587 172.16.8.242_11006 5284
8182291 172.16.8.242_11006 14463
8243584 172.16.8.242_11006 12369
8256573 172.16.8.245_31031 20853
8283387 172.16.8.242_11006 13545

Clustering in Spatio-Temporal
Environment

Clustering

Problem:

To get

List of users online together

Cluster size

How many times?

List of all locations, time etc.

Challenges

Each User has ~100+ sessions

There are ~12000 Users

Only ~5 years to finish your PhD

Basics

Rule#1: Users connected to same Access Point

if(user_x. location .equals(user_y. location )){

}

Rule#2: Users have intersecting time intervals

if(user_x. st <= user_y. et && user_x. et >= user_y. st ){

}

Intersecting Intervals
User A: --------- -------- -------- --------
User B: --------- --------- -------- ----

Min Heap

Sample Output

2 6 14114 14988 17317 24220 29687 29710

2 3 5727 16499 24623

1 3 17798 21831 26081

12 5 15659 15869 19035 24539 24729

4 3 9650 13673 15720

Possible Extensions

Maintain a HeapMap / Vector of locations

Maintain a timestamp schedule they met

Maintain average time they met

Regularity in meeting with this group or other group

FFT (Yibin, Sungwook)

Periodicity of meeting etc.

Markov Chain Markov Models (Jeeyoung)

Min Heap Implementation

First Sort using Location

Then Sort using start time

User minHeap to extract the earliest finish using “End
Time”

Insert into minHeap using “Start Time”

Add all extracted nodes into the HeapMap as key.