Search

Background and aim: Policymakers seek to evaluate their country's scientific performance and measure it in terms of effectiveness and problem-solving. The aim of this study was to make an analytical comparison of Iranian scientific documents in text mining based on domestic and foreign databases. Materials and methods: The present study is descriptive survey with a bibliometric approach. In order to find scientific documents related to text mining in the Scopus database, related terms were searched, and then the results were limited to Iran. Scientific Information Database (SID) was used to search for Persian scientific documents. Bibexcel, VOSviewer, Python programming language, and Excel 2017 were used to analyze the data. Findings: The total number of Iranian scientific documents in text mining in the Scopus citation database was 1082 and 284 (26.25%) of scientific documents indexed in Scopus were in Persian. Moreover, according to the Scientific Information Center, the number of scientific documents in this field was 89 and the number of scientific documents in Persian was 51 (57.30%). The Journal of Lecture Notes in Computer Science has published most international scientific papers in Iran, and the Journal of Signal and Data Processing has published most domestic scientific papers in Iran in text mining. A t-test was used to determine that there was a significant difference in the number of scientific documents in Persian between Scopus and SID databases (p<0.0001). Conclusion: The average growth rate of Iranian scientific documents in text mining was higher than in other subject areas. The United States, Britain, and Australia have had the most collaboration with Iranian researchers in this field. It was also found that international scientific documents in English received more citations than scientific documents in Persian. [ABSTRACT FROM AUTHOR]

As social networks grow, they become more and more complex, and analyzing them becomes complicated. One way to reduce this complexity is to divide the network into subnets, which are also called communities. Dividing social networks into desirable communities can help analysts and experts to understand the behavior and function of the networks. Community detection in networks is a challenging topic in network science and various methods have been proposed for that. Modularity maximization is one of the state-of-the-art methods suggested for community detection. Modularity maximization is an NP-hard problem meaning that no polynomial-time algorithm exists that could solve the problem optimally unless P=NP. A group of approaches that could solve such problems is approximate algorithms. Identifying the influential nodes has many important applications in social networks. This technique could also be used in community detection. To maximize modularity, in this paper, we propose approximate algorithms based on identifying the influential nodes and their influence domain. We used the concept of scale-free networks to prove the approximate factor. Experiments on real-world networks show that the proposed algorithm can compete with the state-of-the-art methods of community detection algorithms. [ABSTRACT FROM AUTHOR]