200 likes | 539 Views
SEEM 5010 A dvanced Database and Information System. Finding a Team of Experts in Social Networks. Theodoros Lappas (UC Riverside) Evimaria Terzi (IBM Almaden) Kun Liu (IBM Almaden). Motivation. A pool of individuals with different skills + A social network Finding a subset
E N D
SEEM 5010 Advanced Database and Information System Finding a Team of Expertsin Social Networks Theodoros Lappas (UC Riverside) Evimaria Terzi (IBM Almaden) Kun Liu (IBM Almaden)
Motivation • A pool of individuals with different skills+ A social network • Finding a subset • TEAM FORMATION Problem • Not only meet Skill Requirements • Can also work Effectively Together as a team How can I find a team of experts that can collaborate effectively in order to complete a given task?
Problem • Given a Task and a set of Experts in Network • Goal: find a set of individuals that can effectively perform the task • Task(T): Set of required skills • Expert(X): Individual with specific skill-set • Network(G): Strength of Relationships • Weights on the edges = Communication Cost
Expertise Networks • Companies (Organizational Structure): • Same group or department • Easily Communicate • Research Community: • Collaboration Networks • Other examples of Social Networks • LinkedIn, Xing and others
How to make a team effective for a given task? • T = {algorithms, software engineering, distributed systems, web programming} • Without considering the social network… • Result: X’ = {A,B,C} • Result: X” = {A,E} A{algorithms} A{algorithms} A{algorithms} B {web programming} B {web programming} C{software engineering, distributed systems} C{software engineering, distributed systems} D{software engineering} E{software engineering, distributed systems, web programming} E{software engineering, distributed systems, web programming}
How to make a team effective for a given task? • With the social network … • TEAM FORMATION with considering a Social Network • Coverage + Communication T={algorithms, software engineering, distributed systems, web programming} A{algorithms} B {web programming} C{software engineering, distributed systems} D{software engineering} E{software engineering, distributed systems, web programming} A, B and C form an effective group to communicate A D A and E could perform task if they could communicate B C E
Problem Definition • Given • The set of n individuals X={1,…,n} • Graph G(X,E) • Task T • Find X’ • With C(X’,T)=T : • And Communication Cost Cc(X’) is minimized • Good Teams • Have all necessary skills • Can communicate effectively E= Edge X’ have the necessary skills
How to measure effective communication? The author proposed two measures: Diameter Minimum Spanning Tree
Diameter (R) • The largest shortest path between any two node in the subgraph • Diameter Communication Cost of X’ • Cc-R(X’) A A D B B C C E E diameter = infty diameter = 1
Minimum Spanning Tree (MST) • The sum of the weights of its edge that spans all the team nodes • MST communication cost • Cc-MST(X’) A A D B B C C E E MST = infty MST = 2
Diameter-TEAM FORMATION problem T={algorithms,java,graphics,PHP} {graphics,PHP,java} {algorithms,graphics} A B A B Skills: algorithms graphics java PHP E E {algorithms,graphics,java} C D {PHP,java} {PHP} Diameter = 2 αrare= PHP Srare={A,C,D} αrare= java Srare={A, C, E} αrare= algorithms Srare={B, E}
Diameter-TEAM FORMATION problem T={algorithms,java,graphics,PHP} {graphics,PHP,java} {algorithms,graphics} A B Skills: algorithms graphics java PHP E E {algorithms,graphics,java} C C D {PHP,java} {PHP} Diameter = 1 αrare= algorithms Srare={B, E} Running time: Quadratic to the number of nodes
MST – TEAM FORMATION problem • The CoverSteiner Algorithm • 2 steps • First step (GreedyCover)… • The social network is ignored and the algorithm focuses on finding a set of individuals X0 • Second step (SteinerTree)… • Find the minimum cost tree that spans all the nodes in X0
MST – TEAM FORMATION problem • SteinerTree problem • Required Vertices • Steiner Vertices • Graph G(X,E) • Set of Required Vertices R • Find G’ sub-graph of G such that G’ contains all the required vertices (R) and MST(G’) is minimized
MST – TEAM FORMATION problem • The EnhancedSteiner Algorithm • EnhanceGraph T={algorithms,java,graphics,PHP} graphics {graphics,PHP,java} {algorithms,graphics} A B java algorithms E E {algorithms,graphics,java} D C D PHP {PHP,java} {PHP} MST Cost = 1
Experiments • Dataset • DBLP dataset • Database, Data Mining, Artificial Intelligence and Theory • Choose about 6000 authors • Skills: Define keywords appearing in paper titles as skills, about 2000 distinct skills • Social Network: Co-Authorship Graph • Tasks: Subsets of keywords with different cardinality
Conclusion • Forming a team of skilled • Minimizing the communication cost • Formulations: • Diameter-TF problem • RarestFirst Algorithm • MST-TF problem • CoverSteiner Algorithm • EnhancedSteiner Algorithm • Qualitative Evaluation