Dinamik ağlar için optimal zaman dilimi tespiti

Extracting a proper dynamic network for modelling a time-dependent complex system is an important issue for analyzing the studied system. Building a correct model is related to finding out critical time points where a system exhibits considerable change. This task is usually executed by a comparative analysis of time series reflecting the features of extracted dynamic networks. These time series are generated by sliding different sizes windows at the entire time span of system. The most commonly used features are the networks' topological properties. In this work, we not only look for topological properties evolution but also propose to measure network stability to detect proper time intervals. We develop three similarity scores to measure node, link, and neighborhood similarities of any consecutive snapshots of a dynamic network. We consult to their statistically expected values which are calculated under the null model of proposed metrics in order to determine whether the network stays stable. We validate the usability of the proposed metrics and of their expected values by demonstrating their reaction to different window sizes. We use two different data sets having different temporal dynamics. First one is an original data set which is log data collected from the Wi-Fi access points in Sabancı University Campus and second one is Enron emails. According to the results, the proposed similarities help to distinguish critical time intervals more effectively than previously proposed metric. Because there is an objective comparison thresholds, i.e. their statistically expected values, proposed similarities are system and methodology independent.