Scholarlink

<p>Identifying influential nodes has attracted the attention of many researchers in recent years. Because of the weak tradeoff between accuracy and running time, and ignoring the community structure by the proposed algorithms in the past research studies, further studies in this area are required. In this paper, we consider communities and also use a novel structure-based neighborhood search to improve exploration strategy of the simulated annealing (SA) algorithm. Moreover, we use the k-shell method for generating a better initial solution instead of random generation. In the proposed algorithm called Ckshell-SA, first, the communities are detected, then the k-shell method is used in each community to find initial candidate nodes locally. Finally, SA algorithm is applied with a neighborhood search that considers the structural properties of the network, and three centralities to find the influential nodes globally. A derivative of the Ckshell-SA method called kshell-SA is also introduced in this paper to examine the impact of considering communities. Unlike the Ckshell-SA, the community structure is neglected, and the k-shell is performed on the whole network in kshell-SA algorithm. Extensive experiments are conducted on eight real-world networks under Independent Cascade Model (IC) and Weighted Independent Cascade Model (WC). The results show that the Ckshell-SA and kshell-SA algorithms outperform the state-of-the-art algorithms concerning influence spread. Furthermore, the results show that Ckshell-SA is more efficient in networks like Facebook with a high Power Law exponent and higher modularity. On the contrary, kshell-SA is more successful in networks like Slashdot or Epinions with lower modularity.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A community-based simulated annealing approach with a new structure-based neighborhood search to identify influential nodes in social networks

  • 期刊论文

  • 发表日期：

    2025年2月17日

查看全文
  • Farzaneh Rajaee Abyaneh,
  • Nasrollah Moghadam Charkari,
  • Mehdy Roayaei

摘要

Identifying influential nodes has attracted the attention of many researchers in recent years. Because of the weak tradeoff between accuracy and running time, and ignoring the community structure by the proposed algorithms in the past research studies, further studies in this area are required. In this paper, we consider communities and also use a novel structure-based neighborhood search to improve exploration strategy of the simulated annealing (SA) algorithm. Moreover, we use the k-shell method for generating a better initial solution instead of random generation. In the proposed algorithm called Ckshell-SA, first, the communities are detected, then the k-shell method is used in each community to find initial candidate nodes locally. Finally, SA algorithm is applied with a neighborhood search that considers the structural properties of the network, and three centralities to find the influential nodes globally. A derivative of the Ckshell-SA method called kshell-SA is also introduced in this paper to examine the impact of considering communities. Unlike the Ckshell-SA, the community structure is neglected, and the k-shell is performed on the whole network in kshell-SA algorithm. Extensive experiments are conducted on eight real-world networks under Independent Cascade Model (IC) and Weighted Independent Cascade Model (WC). The results show that the Ckshell-SA and kshell-SA algorithms outperform the state-of-the-art algorithms concerning influence spread. Furthermore, the results show that Ckshell-SA is more efficient in networks like Facebook with a high Power Law exponent and higher modularity. On the contrary, kshell-SA is more successful in networks like Slashdot or Epinions with lower modularity.