Fuzzy c-means based coincidental link filtering in support of inferring social networks from spatiotemporal data streams

Social network modelling offers an important computational mechanism for analysis of complex system behaviour. Social networks can be established by linking individuals based on observations of certain activities like physical proximity, by exploiting spatiotemporal data streams that have been acquired. A potentially powerful approach to building such networks is to computationally imitate the real-world foraging process of great tits, where the data streams required consist of the times and locations each tit of a certain population has appeared at. The method works by clustering individuals within the population into groups representing different gathering events with respect to the time and location. It links up the individuals appearing in the same events and subsequently, filters out those links that are set up coincidentally. However, the original filtering technique faces significant challenges when considering issues such as time and space complexity and non-unique parameters that are required for removing coincidental links. This paper presents an improved approach by the use of the popular fuzzy c-means method to reinforce the clustering of coincidental links in an emerging social network derived from spatiotemporal data. In particular, all links are organised into two groups: strong links or weak links, prior to the running of the filtering process. The efficacy of the modified version is demonstrated via systematic experimental comparisons against the performance of the original method