Precise Modelling Method for Specific Individual Behavioural Trajectories Based on Community Detection

With the proliferation of location-aware technologies and mobile social platforms, human behavioural trajectory data has experienced explosive growth. How to accurately model the movement patterns of specific individuals has become a hot topic in current research. This paper proposes an innovative framework integrating dynamic community detection with trajectory prediction. By analysing the dynamic association patterns between individuals within social networks, it enhances the accuracy of behavioural trajectory modelling. The methodology first employs dynamic network community detection algorithms to identify groups exhibiting similar behavioural patterns. Subsequently, it combines bidirectional neural temporal processes with graph neural networks to perform spatio-temporal modelling of individual trajectories. Experimental results demonstrate the method's effectiveness across multiple public datasets, showing superior performance to existing benchmark methods in both trajectory prediction accuracy and social plausibility. Specifically, within the TrajNet++ benchmark, it achieves a 7.0% improvement in average accuracy over traditional trajectory prediction models while effectively identifying key behavioural patterns during community evolution. This research offers a novel technical approach for understanding specific individual behaviours within social networks, particularly exploring new methodologies at the intersection of dynamic network analysis and spatio-temporal data modelling.