PhD Proposal

DP
by Diogo Pacheco0 views·Updated Aug 24, 2017
Powered by AI

Need to present on this topic?

Create your own polished presentation in minutes. Just describe what you need.

Free to try, no credit card required

Transcript

18

Clusters based on rivals (whom) - BRA

Football

Clubs as Distributions - BRA

Methodology

Highlight Supporters per State

17

20

Ranking Clubs - BRA

14

Correlation between avg distance from supporters and income.

21

19

Proposed Methodology

Proposed Methodology

Ranking Entities

empathy index

Clusters based on intensity (how) - UK

Characterizing Users by Social Activities

16

apathy index

10

Highlight Entities

13

Ranking Clubs - UK

relative risk

7

mentioned score

Aggregated Characterizations

Supporters as Proxy for Social Disorganization

Analysis of Characterized Elements

elements as distributions

membership

indicator

matrix

clustering

n entities = 20 clubs

users = supporters

mention = @fficial

24

social-economic correlations

11

Clubs as Distributions - UK

(6)

Crime concentration, highlighted clubs, and entropy.

group characterization

group

characterization

matrix

15

8

12

Visual representation of a normalized contingency matrix sorted by preferred entities.

23

Clusters based on similar organs.

Organs Characterization

28

26

Thesis Statement

Organ Donation

Problem Statement

Different Backgrounds

Social activities (posts) in social network sites (SNS) embed useful information not only in the context of conversations (subject), but related to their writers. Online social activities reflect people backgrounds, i.e. we express ourselves online depending on education level, political engagement, economic class, age, race, among others. Therefore, online posts can be used to characterize people in order to learn about the subjects where posts came from and to identify behavior patterns. The characterization helps us to build sensor tracking subjects and providing dynamic information for the population and stakeholders. Moreover, behavior patterns can correlate with some social-economic factors. Ultimately, online posts can be used to dynamically sense social-economic indicator from society.

US states colored according to excessive conversation about specific organs.

27

Clusters of Users

Shopping

29

25

3

2

Remaining Work

Limitations

  • Check amount of data needed
  • Social-economic analysis at city level
  • Create null models from crime analysis

5

Time Table

Football (40%)

Organ (50%)

Language (65%)

  • Perform social-economic analysis on groups of states and users
  • Identify roles in group of users
  • Characterization limited by user's activities
  • Bias from data and types of mention
  • Enough data
  • Subject experts for analysis

Limitations

  • Validate and interpret language characterization results;
  • Characterize countries based on language;
  • Perform social-economic analysis on groups of countries;

30

31

Proposal - Diogo Pacheco

1

What about the past?

2

preferred entity

if "lives in" j

regions

3

Dating

4

m,i users

5

#mentions to j in t

#mentions in t

6

Are online traces enough for us to capture these singularities?

7

Study

8

entity

region

9

n entities = 6 organs

organ donation related people

users =

mention = context + organ

10

n,j entities