چكيده به لاتين
The social network analysis is a powerful diagnostic method for analyzing the nature and pattern of communication between members in terrorist networks, on-line social networks such as Twitter, biological networks, and etc. Changes in social networks may reflect an event or significant behavior within an organization. Detecting these changes can effectively enable early warning and faster response to positive and negative organizational activities. In this paper, communities are randomly formed in a weighted and non-directional network by presenting a model based on Poisson distribution, and the network is monitored. Moreover, the present study investigates how communities are identified with the existing data and an algorithm based on improved modularity has been used for this purpose. Then, using simulation in MATLAB software and analyzing data based on the average run length, the performance of the proposed model was evaluated by applying the SHEWHART control chart, the EWMA and the cumulative sum. The evaluation results show that, if the identification of communities is carried out before monitoring, finding out the source of the change in the network and discovering its cause is much easier. The proper performance of the average run length while using EWMA and CUSUM control charts confirms the validity of the evaluation results in comparison to a sample of studies. In this regard, BONFERRONI’s correction has been implemented on the proposed scheme to provide the necessary factors for comparing the designs.
Keywords: social network monitoring, Poisson distribution, control charts, community detection, average run length
