复杂社交网络中人群搜索团队形成新策略：效率与成本优化

39 浏览量更新于2024-07-15 收藏 654KB PDF 举报

本文探讨了一种新颖的方法，旨在支持在复杂社交网络环境下的人群搜索任务团队组建。该研究发表在《IEEE Transactions on Systems, Man and Cybernetics: Systems》期刊上，以Tan Wenan、上海理工大学计算机与信息技术工程学院的Sun Yong等人作为主要作者，涵盖了商务数据处理、协作工作和自组织控制等多个关键领域。研究的核心内容聚焦于如何有效地解决在大规模社交网络中寻找具有特定技能组合且成本效益高的团队成员的问题。团队组建过程考虑了劳动力成本和沟通成本两个重要因素。通过实验分析，作者展示了他们的方法能够根据专家的技能水平、团队规模以及任务预算动态调整，以达到最优的团队配置。图5展示了专家按照劳动成本进行的分组情况，这表明方法能够在众多参与者中识别出最经济高效的组合。随后的图6和图7比较了不同技能数量（#skill=2和#skill=6）下团队的劳动力成本，清晰地展示了随着技能需求增加，团队成本的变化趋势。权重和预算对团队成本的影响也在研究中得到了深入探究。图8展示了随着权重参数的变化，如何影响劳动力成本的分布，而图9则揭示了预算对沟通成本的影响。图10则进一步展示了在有限预算下，团队之间的通信成本是如何平衡的。此外，研究还评估了算法在实际数据集上的性能，包括DBLP（一个知名的计算机科学文献数据库）和一个反相关数据集。图17和图18分别展示了在这些数据集上的运行时间，结果显示了方法在处理大规模和复杂网络时的有效性和效率。这篇研究论文提供了一个创新的框架，用于优化复杂社交网络中人群搜索任务的团队组建策略，通过量化成本、技能匹配和效率，为在线协作平台和企业寻找最佳团队合作模式提供了有价值的技术支持。其结果对于提升协作项目的成功率、降低运营成本以及提高团队生产力具有重要的实践意义。

the costs on experts and interactions were weighted and

summed as input values of edges in a social network, and the

heuristic algorithm was adopted to update and evolve team

schemes iteratively until the termination conditions for the best

scheme were satisfied (Kargar et al. 2012, 2013a, b). However,

their model did not cope with the issue of unifying the

cost-related quantities since they depend on the scope of

anticipated properties of team candidates. In addition, they

failed to give two cost utility functions. Reexamining the

example in Fig. 1, the sum of expert cost of team A is 70 Yuan;

while the interaction cost of the team is 0.1. The summation of

expert cost and interaction cost became pointless if the

unification or normalization of quantities were not performed.

The team formation problem will become very challenging

when the pool of expert candidates is large in a social network, it

takes intensive computation in evaluating a large number of

team candidates to update and evolve better team schemes.

To solve the aforementioned problems in existing team

formation models, we propose a new method called the

Skyline-based Team Formation (SkylineTF), where the notion

of Skyline is adopted in the team formation model. Firstly, a

social network is analyzed by the method of betweenness

centrality to select seeds of experts, and the utility functions are

used to convert expert quality vectors into real values. Secondly,

the notions of Skyline are adopted to obtain expert candidates

that are not dominated by other experts. Thirdly, the weights of

cost aggregate functions are adjusted iteratively until the expert

team is formed according to the termination condition of

optimization. Finally, we conduct the experiment to illustrate

the effectiveness of the proposed method in comparison with

existing models in sense that the searching time for expert

candidates has been reduced remarkably.

The rest of the paper is organized as follows. In Section II, we

discuss the relevant work on the methodologies for team

formation. In Section III, the concept of the team discovery

model is defined and discussed. In Section IV, a new team

discovery model is proposed for crowdsourcing tasks. In

Section V, the new algorithm is developed to assign tasks to an

expert team in a social network automatically. In Section VI, the

experiment is reported as the validation of the proposed method.

Finally, a summary of the presented work is provided with a

brief discussion of future work in Section VII.

II. R

ELATED

ORKS

In this work, we focus on the discovery of an expert team in a

social network to accomplish a specified project. The success of

the project is measured by the effectiveness and engagement

levels of team members in collaboration and communication. In

other words, the primary objective of the team formation is to

minimize the operation cost of the developed team in fulfilling

the specified business missions. Therefore, it is crucial to take

into account of the effectiveness of collaborations in identifying

team members. The pioneer work in developing an expert team

from a social network was made Lappas et al. (2009); they

implemented two algorithms form expert teams based on an

analysis of existing relations. Li et al. (2013) generalized the

team formation problem in sense that every required skill is

associated with a given number of experts; their goal was to

improve the effectiveness and efficiency of the developed team.

Anagnostopoulos et al. (2012) also discussed the factor of the

balance of experts’ workload, and took into account in the team

formation. In the method proposed by Majumder et al. (2012),

the criterion that no expert member took an overload was placed

as a constraint in identifying team members. In the

aforementioned algorithms (Lappas et al. 2009,

Anagnostopoulos et al. 2010, Anagnostopoulos et al. 2012,

Datta et al. 2012, Li et al. 2013), the effectiveness of

collaboration was represented by either of two cost functions,

i.e., diameter communication cost and minimum spanning tree

cost.

The roles of participators in a team are not necessary the

same. For some projects, the project team requires to specify a

team leader who is in charge of overseeing the activities and

coordinating team members for the project (Kargar and An

2011). In such scenarios, each team member needs to

communicate with the project leader directly to update the

progresses of the ongoing activities; the team leader is

responsible of performing its decision-making activities to

guide the members for effective collaboration at the project

level. However, neither of the diameter communication cost or

minimum spanning tree cost is appropriate under such

circumstances.

To solve this problem, Kargar and An (2011) defined a

project leader who was in charge of coordinating activities in

their team formation. New cost function type called the leader

cost function is defined to measure the cost of the

communication of the leader with associated members. In

optimizing the selection of the leader and associated team

members, Kargar et al. (2013a, b) used the brute-force

algorithm to evaluate each potential member as a candidate

leader and identify its team member with the leader. Therefore,

the best team was defined by evaluating and comparing all the

candidates in the social network exhaustively; the brute-force

algorithm required intensive computations when the number of

qualified experts in the network is large. To reduce computation

in team formation, Juang et al. (2013) selected a member with

multiple shortest paths as the leader candidate for a lower

collaboration cost. Since the communication cost was measured

by the geographic distances of the leader and team members, the

concept of the betweenness centrality was used as the selection

criterion for a project leader. Their experiments illustrated the

improved algorithm generated better outcomes.

All of the above algorithms were developed to form project

teams from social networks. However, the common drawbacks

of these algorithms were the lack of considerations on labor

costs of experts. In the real-world business, most of expert

services are available by pay for services, and service rates vary

from one expert to another even for the same nature of

responsibility in project. Therefore, the cost evaluation merely

on team interactions is insufficient. It is desirable that the expert

team discovery model take into consideration of both of labor

cost and interaction cost of the team. Some preliminary works

by Kargar et al. (2012, 2013a, b) took a weighted summation of

labor cost and interaction cost as the attributes of edges in the

graphic representation of team in a social network; they

suggested to use a heuristic algorithm to optimize the team

Page 3 of 15

剩余15页未读，继续阅读

weixin_38689338

粉丝: 9
资源: 974

复杂社交网络中人群搜索团队形成新策略：效率与成本优化

一种将工作流任务众包到社交网络的高效算法

选修七单词检测.doc

软件项目中的人员管理.ppt

字符串与日期时间处理：结合String库的高效方法，优化时间管理技巧

信用风险控制中的机器学习算法比较分析

敏捷开发中的用户故事：如何编写、管理和优化

基于Vue的智能配方推荐系统.js+SpringBoot+MySQL开发，高分成品毕业设

m5-forecasting-accuracy数据集

kafka 18道面试题和答案（2025最新版）.docx

小程序毕业设计-基于Springboot的农场管理微信小程序【代码+论文】

最新资源