Temporal Network Pattern Identification by Community Modelling

Abstract

Temporal network mining tasks are usually hard problems. This is because we need to face not only a large amount of data but also its non-stationary nature. In this paper, we propose a method for temporal network pattern representation and pattern change detection following the reductionist approach. The main idea is to model each stable (durable) state of a given temporal network as a community in a sampled static network and the temporal state change is represented by the transition from one community to another. For this purpose, a reduced static single-layer network, called a target network, is constructed by sampling and rearranging the original temporal network. Our approach provides a general way not only for temporal networks but also for data stream mining in topological space. Simulation results on artificial and real temporal networks show that the proposed method can group different temporal states into different communities with a very reduced amount of sampled nodes. In real-world applications, data often arrives in streams and must be analyzed in real-time. Patterns and relations in such data are usually not stable but evolve over time 1,2. In dynamical and non-stationary environments, the data distribution can change, yielding the phenomenon of concept drift 2,3. One example is the stock market, where the massive data exchange is strongly related to macro and micro political events, economic events and other factors, such as natural and man-made disasters. Another example is the pattern of customers†™ buying preferences, which may depend on the season, availability, inflation rate, etc. Some common types of changes may include a gradual change over time, a recurring or cyclical change, and a sudden or abrupt change. Different concept drift detection and handling schemes may be required for each situation. In this context, uncovering data stream patterns is a fundamental task not only to detect concept drift but also to predict future behaviors. Recently, some works have employed networks as a way for representing data from many real systems 4-7 , leading to complex network research. Complex network refers to large scale graphs with nontrivial connection patterns 8-11. One of the most important features of a complex network is the presence of communities. Detecting communities in these systems has become a fundamental task to help us understand how local patterns (represented by sub-networks) interact and produce global behavior. From a network topology point of view, a community is a sub-graph in which the inner links are dense while the outer connections are relatively sparse 4,12,13. In terms of the information-theoretic field, a community is seen as the module that can diminish or retard the propagation flow of information in the system, for a considerable period of time 14-16. In the machine learning domain, community detection provides new unsupervised learning methods, where each community corresponds to a data cluster in the clustering problem 4. Large systems usually are high-dimensional...

New Cross-Talks between Pathways Involved in Grapevine Infection with ‘Candidatus Phytoplasma solani’ Revealed by Temporal Network Modelling

“…An alternative approach to using the dissipation index can include temporal community detection [40]; however, this method is not necessarily suitable for the data-scarce regime with only two time points considered in this work. The indication that the post-hoc analysis of communities is more suitable for analysis was also recently demonstrated when studying fire events in a portion of the Amazon basin [41].…”

Dynamic Community Detection into Analyzing of Wildfires Events

“…The authors showed the effectivity of the proposed method in some artificial datasets, and in a case study analyzing wildfire events from the same temporal Chronnets dataset [2,13]. As a result, they also detected two central communities in the Amazon region, each one corresponding to different periods of the year: the south-hemisphere winter season, with a high tendency of fires, and the south-hemisphere summer season, with a low frequency of fires [14]. However, these communities represent the global state of the wildfire system in the Amazon basin, not the micro spatialtemporal particularities or patterns into the microregions.…”

“…In this work, we aim to analyze the temporal information stored on the wildfire dataset by using temporal chronnets. Our analysis differs from previous works [2,13,14] once we seek to validate the results of the temporal community detection methods when modeling the wildfires, mainly concerning its spatial incidence and extension. Our methodology can reveal where and how often specific fire event patterns occur over the years.…”

“…Recently, Gao et al [14] proposed a method for mining temporal networks by representative nodes in a community identification approach. They tackled the problem from the change detection perspective, where each stable temporal state of the dataset is represented as a community.…”

Static and Dynamic Community Detection Methods That Optimize a Specific Objective Function: A Survey and Experimental Evaluation

“…Xubo et al [139] proposed a block-based method for detecting community structure from temporal networks, whose data can change or evolve over time. The method employs a reduction strategy using sampling.…”