Improvement Of Social Network Change Detection Based On Time Series Analysis

Social Network Change Detection (SNCD) is a rising research field, whichintegrated several different subjects, such as sociology and statistics. Deeply understandthe change of social network with the help of social network change detection has animportant meaning to against the sudden activities.As there are so many different features between measure of social network andmeasure of production process, there may make some false alarms if one make SPCquality control based on the data without preprocess, which will decrease the controlaccuracy. As a result, this thesis will analysis the condition of tendency、periodicity andautocorrelation, and provide a method based on time series analysis to delete thesefactors to satisfied the requirement of SPC, then control and detect changes on thestatistical data with traditional control chart and CUSUM control chart. Analyze thechange detection results with the help of network organization event details, one couldunderstand the social network change detection result more accurately, and determinethe network state and change reason of social network changes.Enron Email Dataset is a database that records the Email connections betweenemployees before Enron bankrupt. This article will construct a social network with theEnron Email Dataset, gather the betweenness、closeness and degree over nearly3years,and check the changes. With the comparison of experiment results and real events inEnron, one can get the conclusion that this method is effective.