A parallel team formation approach using crowd intelligence from social network

Abstract

- In the recent year, with the emergence of various smart devices, the data is explosively increasing in the social Internet of Things (IoT) such as human healthcare. These data mainly involve information about user behaviors collected from various heterogeneous wireless sensor and social networks. Therefore, it is vital to analyze the data to find hidden meaning and convert it into valuable information. Due to the lack of capability to handle a wide range of queries, a traditional relational database provides inefficient analysis for the data. A graph database can easily store and analyze the data from various heterogeneous wireless sensor and social network using team formation algorithm. In the healthcare field, it is important to form a team that manages patients' health efficiently. The final goal of team formation is to organize experts who can perform task of data analysis. However, the existing team formation algorithms rely on a centralized computing environment and require high communication cost among experts to form a team. In this paper, we propose a parallel team formation method on apache spark (PTFS) to analyze graph data considering the crowd intelligence capability that exists in the graph data and social network. The PTFS employs two computation stages - a find skill and a merger subgraph and provides the parallel execution of many map tasks of graph data analysis. The experimental evaluation of the proposed method on a graph dataset demonstrates that it minimizes the communicating cost of the team members to form an optimized expert team in which a desired skill set is assigned to accomplish the graph data analysis.