46
The Anatomy of Developer Social Networks Qiaona HONG Supervisor: Prof. Shing-Chi Cheung 1

The Anatomy of Developer Social Networks

Embed Size (px)

DESCRIPTION

Qiaona hong MPhil defense slides

Citation preview

Page 1: The Anatomy of Developer Social Networks

1

The Anatomy of Developer Social Networks

Qiaona HONGSupervisor: Prof. Shing-Chi Cheung

Page 2: The Anatomy of Developer Social Networks

2

Social Network• Study the Topological Structure of Social

Network– Y. Y. Ahn @WWW '07; A. Mislove@IMC '07

• Study the Community Structure of Social Network– V. D. Blondel@ Journal of Statistical Mechanics:

Theory and Experiment; Y. R. Lin@WI '07• Techniques to visualize the social network

– Jeffrey Heer@InfoVis '05• Influential People & Information Diffusion

– Kimura, M.@InfoVis '07• Friend Recommendation

– Nitai B. Silva@WCCI‘10

General Social Network (GSN)

Page 3: The Anatomy of Developer Social Networks

3

Research Questions

• Q1: What are the similarities and differences between DSNs and GSNs?

Page 4: The Anatomy of Developer Social Networks

4

Research Questions

• Q1: What are the similarities and differences between DSNs and GSNs?

• Q2: How do DSNs evolve over time?• Q3: How do communities evolve in DSNs?• Q4: What are the similarities and differences

between DSNs extracted using different social linkage indicators?

Page 5: The Anatomy of Developer Social Networks

5

Research Questions

• Q1: What are the similarities and differences between DSNs and GSNs?

• Q2: How do DSNs evolve over time?• Q3: How do communities evolve in DSNs?• Q4: What are the similarities and differences

between DSNs extracted using different social linkage indicators?

Qiaona HONG, Sunghun Kim, S.C. Cheung and Christian Bird, “Understanding a Developer Social Network and its Evolution”, in Proceedings of the 27th IEEE International Conference on Software Maintenance, 2011.

Page 6: The Anatomy of Developer Social Networks

6

Subjects

• Mozilla Bug Report: 2000-2009 – 496,692 bug reports– 3,893,025 comments

• Mozilla CVS Log: 2000-2009– 44394 revisions

• Eclipse Bug Report: 2002-2009– 294,938 bug reports– 1,618,667 comments

• Eclipse CVS Log: 2002-2009– 22493 revisions

Page 7: The Anatomy of Developer Social Networks

7

DSN Extraction Approach

David Comment 1

Bug Report 1

Jack Comment 3

Bob Comment 2

Bob Comment 1

Bug Report 2

Bill Comment 3

Jack Comment 2

Bug Report 4

Jack Comment 3

Bob Comment 2David Comment 1

Bug Report 3

Bill Comment 3

Bob Comment 2

Jack Comment 3

David Bill

Bob Jack

Page 8: The Anatomy of Developer Social Networks

8

DSN Extraction Approach

David Comment 1

Bug Report 1

Jack Comment 3

Bob Comment 2

Bob Comment 1

Bug Report 2

Bill Comment 3

Jack Comment 2

Bug Report 4

Jack Comment 3

Bob Comment 2David Comment 1

Bug Report 3

Bill Comment 3

Bob Comment 2

Jack Comment 3

David Bill

Bob Jack

1

2 2

4

2 2

Page 9: The Anatomy of Developer Social Networks

9

DSN Extraction Approach

David Comment 1

Bug Report 1

Jack Comment 3

Bob Comment 2

Bob Comment 1

Bug Report 2

Bill Comment 3

Jack Comment 2

Bug Report 4

Jack Comment 3

Bob Comment 2David Comment 1

Bug Report 3

Bill Comment 3

Bob Comment 2

Jack Comment 3

David Bill

Bob Jack

4

Page 10: The Anatomy of Developer Social Networks

10

DSN Extraction Approach

David Comment 1

Bug Report 1

Jack Comment 3

Bob Comment 2

Bob Comment 1

Bug Report 2

Bill Comment 3

Jack Comment 2

Bug Report 4

Jack Comment 3

Bob Comment 2David Comment 1

Bug Report 3

Bill Comment 3

Bob Comment 2

Jack Comment 3

Bob Jack

Page 11: The Anatomy of Developer Social Networks

11

Metrics

• Degree Distribution– The number of edges connected to a node

• Degree of Separation– The shortest path between two nodes

• Modularity– To measure the quality of division of nodes

• Community Size– The number of nodes within a community

Page 12: The Anatomy of Developer Social Networks

12

Modularity

A B0.51 0.176

• According to A. Clauset’s work, modularity of 0.3 is a good indicator of significant community structure in a network

• When the modularity is 0, the community structure is no stronger than that of a randomly generated network

Page 13: The Anatomy of Developer Social Networks

13

Communities in DSN

• Identified Communities in DSN– Louvain Algorithm (by optimizing modularity)– 50 different input ordering of nodes

Page 14: The Anatomy of Developer Social Networks

14

?Q1: What are the similarities

and differences between DSNs and GSNs

Degree of Distribution Degree of Separation

Modularity Community Size

Page 15: The Anatomy of Developer Social Networks

15

Degree Distribution

(1) MozillaDSN-BR (2) MozillaDSN-CL

(3) EclipseDSN-BR (4) EclipseDSN-CL

Q1: What are the similarities and differences between DSNs and GSNs

Page 16: The Anatomy of Developer Social Networks

16

Degree Distribution

(1) MozillaDSN-BR (2) MozillaDSN-CL

(3) EclipseDSN-BR (4) EclipseDSN-CL

Q1: What are the similarities and differences between DSNs and GSNs

Page 17: The Anatomy of Developer Social Networks

17

Degree Distribution

(1) MozillaDSN-BR (2) MozillaDSN-CL

(3) EclipseDSN-BR (4) EclipseDSN-CL

• Quantitative power law fit test– An approach of analyzing power law distributed

data introduced by A. Clauset et al.

• P-value : The likelihood that degree distribution does actually follow a power-law– If p-value is less than 0.1, the power law is

rejected.

Q1: What are the similarities and differences between DSNs and GSNs

Page 18: The Anatomy of Developer Social Networks

18

Degree Distribution

(1) MozillaDSN-BR (2) MozillaDSN-CL

(3) EclipseDSN-BR (4) EclipseDSN-CL

P-value<0.1 some<0.1,other>0.1

Different from GSNs, DSNs do not follow power-law

Q1: What are the similarities and differences between DSNs and GSNs

Page 19: The Anatomy of Developer Social Networks

19

Degree of Separation

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of SeparationP

roba

bilit

y

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of SeparationP

roba

bilit

y

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 180.

00.

20.

40.

60.

81.

0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of SeparationP

roba

bilit

y

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

Q1: What are the similarities and differences between DSNs and GSNs

Page 20: The Anatomy of Developer Social Networks

20

Degree of Separation

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of SeparationP

roba

bilit

y

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of SeparationP

roba

bilit

y

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 180.

00.

20.

40.

60.

81.

0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of SeparationP

roba

bilit

y

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

4.12

90% (6)

Q1: What are the similarities and differences between DSNs and GSNs

Page 21: The Anatomy of Developer Social Networks

21

Degree of Separation

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of SeparationP

roba

bilit

y

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of SeparationP

roba

bilit

y

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 180.

00.

20.

40.

60.

81.

0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of SeparationP

roba

bilit

y

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

MozillaDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-CL

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-BR

0 2 4 6 8 10 12 14 16 18

0.0

0.2

0.4

0.6

EclipseDSN-CL

Degree of Separation

Pro

babi

lity

Distance between two developers

1-month DSN3-month DSN6-month DSN

1-year DSN2-year DSN4-year DSN

tw itter(8000 sample)cyw orld(3000 sample)

3.0 2.1

4.0 2.5

Developers in DSN are much closer to each other than participants in GSN.

Q1: What are the similarities and differences between DSNs and GSNs

Page 22: The Anatomy of Developer Social Networks

22

Modularity

Network

Mod

ular

ity

0.3

0.4

0.5

0.6

0.7EclipseDSN-BR

0.3

0.4

0.5

0.6

0.7

EclipseDSN-CL

0.3

0.4

0.5

0.6

0.7

MozillaDSN-BR

0.3

0.4

0.5

0.6

0.7MozillaDSN-CL

Modularity

Similar to GSNs, all DSNs have significant community structure

Q1: What are the similarities and differences between DSNs and GSNs

Page 23: The Anatomy of Developer Social Networks

23

Community Size

(1) MozillaDSN-BR (2) MozillaDSN-CL

(3) EclipseDSN-BR (4) EclipseDSN-CL

Q1: What are the similarities and differences between DSNs and GSNs

Page 24: The Anatomy of Developer Social Networks

24

Community Size

(1) MozillaDSN-BR (2) MozillaDSN-CL

(3) EclipseDSN-BR (4) EclipseDSN-CL

28%

Q1: What are the similarities and differences between DSNs and GSNs

Page 25: The Anatomy of Developer Social Networks

25

Community Size

(1) MozillaDSN-BR (2) MozillaDSN-CL

(3) EclipseDSN-BR (4) EclipseDSN-CL

21%-36% 23%-43%

15%-30% 23%-33%

Q1: What are the similarities and differences between DSNs and GSNs

Page 26: The Anatomy of Developer Social Networks

26

?Q4:What are the similarities and differences between DSNs extracted using different social linkage indicators

Q2: How do DSNs evolve over time?

Degree of Distribution Degree of Separation

Modularity Community Size

Page 27: The Anatomy of Developer Social Networks

27

Change of Developer Size

DSNs-BR always have more developers than DSNs-CL

Q2: How do DSNs evolve over time?

Page 28: The Anatomy of Developer Social Networks

28

Change of Percentage of New Comers

DSNs-BR always have higher percentage of new comers than DSNs-CL

Q2: How do DSNs evolve over time?

Page 29: The Anatomy of Developer Social Networks

29

Change of Degree Distribution

(1) MozillaDSN-BR (2) MozillaDSN-CL

(3) EclipseDSN-BR (4) EclipseDSN-CL

Cumulative Degree Distribution Evolution

Degree

Pro

ba

bili

ty(d

eg

ree

>=

x)

100

101

102

103

104

10-4

10-3

10-2

10-1

100

2000fh2000sh2001fh2001sh2002fh2002sh2003fh2003sh2004fh2004sh2005fh2005sh2006fh2006sh2007fh2007sh2008fh2008sh2009fh2009sh

Cumulative Degree Distribution Evolution

Degree

Pro

ba

bili

ty(d

eg

ree

>=

x)

100

101

102

103

104

10-4

10-3

10-2

10-1

100

2000fh2000sh2001fh2001sh2002fh2002sh2003fh2003sh2004fh2004sh2005fh2005sh2006fh2006sh2007fh2007sh2008fh2008sh2009fh2009sh

Cumulative Degree Distribution Evolution

Degree

Pro

ba

bili

ty(d

eg

ree

>=

x)

100

101

102

103

104

10-4

10-3

10-2

10-1

100

2002fh2002sh2003fh2003sh2004fh2004sh2005fh2005sh2006fh2006sh2007fh2007sh2008fh2008sh2009fh2009sh

Q2: How do DSNs evolve over time?

Page 30: The Anatomy of Developer Social Networks

30

Change of Degree Distribution

(1) MozillaDSN-BR (2) MozillaDSN-CL

(3) EclipseDSN-BR (4) EclipseDSN-CL

Cumulative Degree Distribution Evolution

Degree

Pro

ba

bili

ty(d

eg

ree

>=

x)

100

101

102

103

104

10-4

10-3

10-2

10-1

100

2000fh2000sh2001fh2001sh2002fh2002sh2003fh2003sh2004fh2004sh2005fh2005sh2006fh2006sh2007fh2007sh2008fh2008sh2009fh2009sh

Cumulative Degree Distribution Evolution

Degree

Pro

ba

bili

ty(d

eg

ree

>=

x)

100

101

102

103

104

10-4

10-3

10-2

10-1

100

2000fh2000sh2001fh2001sh2002fh2002sh2003fh2003sh2004fh2004sh2005fh2005sh2006fh2006sh2007fh2007sh2008fh2008sh2009fh2009sh

Cumulative Degree Distribution Evolution

Degree

Pro

ba

bili

ty(d

eg

ree

>=

x)

100

101

102

103

104

10-4

10-3

10-2

10-1

100

2002fh2002sh2003fh2003sh2004fh2004sh2005fh2005sh2006fh2006sh2007fh2007sh2008fh2008sh2009fh2009sh

14(20) DSNs <0.1

15(16) DSNs <0.1

DSNs which has p-value <0.1 have

bigger size

DSNs do not follow power law degree distribution over time

Q2: How do DSNs evolve over time?

Page 31: The Anatomy of Developer Social Networks

31

Change of Degree of Separation

0 3 6 9 12 0 3 6 9 12 0 3 6 9 12 0 3 6 9 12

Degree of Separation Evolution

Distance between two developers

Pro

ba

bili

ty

0.0

0.2

0.4

0.6

0.8

1.0

MozillaDSN-BR MozillaDSN-CL EclipseDSN-BR EclipseDSN-CL

2000fh2000sh2001fh2001sh

2002fh2002sh2003fh2003sh

2004fh2004sh2005fh2005sh

2006fh2006sh2007fh2007sh

2008fh2008sh2009fh2009sh

2.45-3.02 2.21-3.91 2.54-3.97 2.22-3.71

Developers in DSN are always closer to each other over time

Q2: How do DSNs evolve over time?

Page 32: The Anatomy of Developer Social Networks

32

Modularity Evolution

Time

Mod

ular

ity

0.2

0.4

0.6

2000

f

2000

s

2001

f

2001

s

2002

f

2002

s

2003

f

2003

s

2004

f

2004

s

2005

f

2005

s

2006

f

2006

s

2007

f

2007

s

2008

f

2008

s

2009

f

2009

s

EclipseDSN-BR

0.2

0.4

0.6

EclipseDSN-CL

0.2

0.4

0.6

MozillaDSN-BR

0.2

0.4

0.6

MozillaDSN-CL

Modularity

There are clear increases in modularity over time for MozillaDSN-CL, MozillaDSN-BR, EclipseDSN-BR

Q2: How do DSNs evolve over time?

Page 33: The Anatomy of Developer Social Networks

33

Modularity Evolution

Time

Mod

ular

ity

0.2

0.4

0.6

2000

f

2000

s

2001

f

2001

s

2002

f

2002

s

2003

f

2003

s

2004

f

2004

s

2005

f

2005

s

2006

f

2006

s

2007

f

2007

s

2008

f

2008

s

2009

f

2009

s

EclipseDSN-BR

0.2

0.4

0.6

EclipseDSN-CL

0.2

0.4

0.6

MozillaDSN-BR

0.2

0.4

0.6

MozillaDSN-CL

Modularity

0.3

0.3

0.3

0.3

All DSNs tend to have stronger community structure over time

Q2: How do DSNs evolve over time?

Page 34: The Anatomy of Developer Social Networks

34

Change of Community Size

(1) MozillaDSN-BR (2) MozillaDSN-CL

(3) EclipseDSN-BR (4) EclipseDSN-CL

Q2: How do DSNs evolve over time?

Page 35: The Anatomy of Developer Social Networks

35

Change of Community Size

(1) MozillaDSN-BR (2) MozillaDSN-CL

(3) EclipseDSN-BR (4) EclipseDSN-CL

Q2: How do DSNs evolve over time?

Page 36: The Anatomy of Developer Social Networks

36

? Q3:How do communities evolve in DSNs?

Page 37: The Anatomy of Developer Social Networks

Six patterns

Expand Shrink

Merge Split

Extinct Emerge

Q3:How do communities evolve in DSNs?

Page 38: The Anatomy of Developer Social Networks

Community evolution map

Mozilla

Eclipse

Q3:How do communities evolve in DSNs?

Page 39: The Anatomy of Developer Social Networks

Community evolution map

Mozilla

Eclipse

FireFox Web Tools (WTP) Project

Q3:How do communities evolve in DSNs?

Page 40: The Anatomy of Developer Social Networks

Community evolution map

Firefox 3 release

FireFox

Q3:How do communities evolve in DSNs?

Page 41: The Anatomy of Developer Social Networks

41

Findings

• Comparison– DSN do not follow power law degree distribution– Developers in DSN are closer to each than people in GSN– DSN has smaller community size– Similar to GSNs, DSN has significant community structure– DSNs-BR have a higher percentage of new comers– Developers in DSNs-CL are closer

• Evolution– Over time DSN tends to have stronger and more stable

community structure– Evolution patterns reflect events happened

Page 42: The Anatomy of Developer Social Networks

42

Discussion and Future Work

• Discussion– Our conclusions are only based on Mozilla and

Eclipse• Our approach could be easily applied to other projects

– Use threshold to remove casual users• Their low connectivity implies that these participants

have little impact on the properties of DSNs that we have examined

Page 43: The Anatomy of Developer Social Networks

43

Discussion and Future Work

• Future Work– Apply social analysis to other open source

software development process• Peer review processes

– Xin Yang, Raula Gaikovina Kula, Camargo Cruz Ana Erika, Norihiro Yoshida, Kazuki Hamasaki, Kenji Fujiwara, and Hajimu Iida, "Understanding OSS Peer Review Roles in Peer Review Social Network (PeRSoN)," In Proceedings of the 19th Asia-Pacific Software Engineering Conference (APSEC2012)

Page 44: The Anatomy of Developer Social Networks

44

Related Work

• C. Bird, A. Gourley, P. Devanbu, M. Gertz, and A. Swaminathan, "Mining email social networks," in MSR '06: Proceedings of the 2006 international workshop on Mining software repositories. New York, NY, USA: ACM Press, 2006, pp. 137-143.

• C. Bird, D. Pattison, R. D'Souza, V. Filkov, and P. Devanbu, "Latent social structure in open source projects," in Proceedings of the 16th ACM SIGSOFT International Symposium on Foundations of software engineering, ser. SIGSOFT '08/FSE-16. New York, NY, USA: ACM, 2008, pp. 24-35.

• A. Meneely, and L. Williams, "Secure Open Source Collaboration: An Empirical Study of Linus’ Law " in Computer and Communications Security, Chicago, IL, 2009, p. 453-462.

Page 45: The Anatomy of Developer Social Networks

45

Acknowledgements

• Prof. Shing-Chi CHEUNG• Prof. Sunghun Kim• Dr. Christian Bird

Page 46: The Anatomy of Developer Social Networks

46

Thank YouQ&A