Finding ‘‘interesting’’ trends in social networks using frequent pattern
description
Transcript of Finding ‘‘interesting’’ trends in social networks using frequent pattern
Intelligent Database Systems Lab
Presenter : MIN-CONG WUAuthors : PUTERI N.E. NOHUDDIN , FRANS COENEN , ROB CHRISTLEY , CHRISTIAN SETZKORN , YOGESH PATEL , SHANE WILLIAMS C
2012.KBS
Finding ‘‘interesting’’ trends in social networks using frequent pattern mining and self organizing maps
Intelligent Database Systems Lab
Outlines
MotivationObjectivesMethodologyExperimentsConclusionsComments
Intelligent Database Systems Lab
Motivation• Number of trends may be identified, too many
to allow simple inspection by decision makers. Some mechanism was therefore required to allow the simple presentation of trend lines.
Intelligent Database Systems Lab
Objectives
• Generating frequent pattern trends,and use SOM technology a process for assisting the analysis of the identified trends, and to identify ‘‘interesting’’ changes in trends.
Intelligent Database Systems Lab
Methodology-The trend mining mechanism
Intelligent Database Systems Lab
Methodology - Frequent pattern trend mining (TM-TFP)Input: Data set :{t1,t2,..,tn}, ti={a,…,z}a={a1,a2,…,an}, support:3Interestpattern: {a,c,s}Example:support:3Interestpattern: {a,c,s}ID Item set ordered
Intelligent Database Systems Lab
id
m
m
IDConditions Target
Conditions Targettree
Methodology - Frequent pattern trend mining (TM-TFP)
Frequentpattern
{a,b,c,d}={0,0,2500,3311,2718,0,0,0,2779}{a,b,c,e}={3,12,6,0,100,2437,0,56,79}{a,c,e,f}={0,0,0,2568,345,23,90,0,459}
Intelligent Database Systems Lab
Methodology – Trend clusteringInput: v1,v1,..,vnProcess: || V – Wi || = min { || V – Wj || }
Output: BMU
CE/通用格式CE/通用格式
CE/通用格式
CE/通用格式
CE/通用格式CE/通用格式
CE/通用格式
CE/通用格式
Intelligent Database Systems Lab
Methodology – Trend clusters analysis
*
e*k
*
Intelligent Database Systems Lab
Experiment - Cattle movement database
Intelligent Database Systems Lab
Experiment - Cattle movement trend mining
Intelligent Database Systems Lab
Experiment - Deeside Insurance database
Intelligent Database Systems Lab
Experiment - Deeside Insurance trend mining
Intelligent Database Systems Lab
Conclusions
• By employing the SOM clustering technique, the large number of trend lines that are typically identified may be grouped to facilitate a better understanding of the nature of the trends.
Intelligent Database Systems Lab
Comments• Advantages
- a better understanding of the nature of the trends.Applications- self organizing map