Special Issue on Trajectory-based Behaviour Analytics

The Fourth Paradigm, namely Data Driven Scientific Discovery, has already become an important emerging approach to realize potentials from massive data sets in the Big Data era. It impacts many domains ranging from environmental to social; it also drives new directions in data analytics research. Very often, these research directions are multi-disciplinary in nature, and they involve not only Computer Science techniques, but also incorporate domain knowledge in the analysis. To explore new value creation from Big Data, one important data source is the trajectory information of entities (such as animals, vehicles, and humans) monitored and captured in real-time. This information naturally reveals the details of instantaneous transactional behaviours conducted by entities, which is closely related to the complex interior behaviours in the form of multiple multivariate time series data with varied locations. This forms the need and emergence of behaviour modelling (i.e. understanding behaviours from cognitive and analytics perspectives) and behaviour analytics system construction (i.e. developing cognition-as-a-service systems to support decision making). Performing real-time behaviour analytics faces many new challenges. Besides the massive volume of data and the consideration of real-time performance and return-on-investment, the complication is often due to the nature of the multiple data sources involved – they are heterogeneous, inter-dependent, and their data quality such as accuracy might not be what we hope to have. Another complication is the domain knowledge – how to couple domain knowledge in the data source selection process as well as how to use them in the analysis more effectively and efficiently. This special issue offers an updated overview of the research field in line with the trajectory-based behaviour analytics. It was generated from the First International Workshop on Trajectory-based Behaviour Analytics, which was part of the Twenty-Ninth AAAI Conference on Artificial Intelligence 2015. A total of sixteen papers were submitted, among which four are selected for this special issue. The first paper, “Identifying locations from geospatial trajectories”, by Thomason and Griffiths et al. addresses the problem of extracting significant locations from the trajectory data for effective and efficient behaviour analysis and prediction. The second paper, “Discovery of stop regions for understanding repeat travel behaviors of moving objects”, by Huang and Je et al. goes one step further and segments trajectory data traces into regions, each of which will be summarized using attributes such as “stay stability” for later mining process. The third paper, “Self-regularized causal structure discovery for trajectory-based networks”, by Chu and Wong et al. addresses the heterogeneity and human cognitive aspects of trajectory data mining using the notion of causal time-varying dynamic Bayesian network. Finally, the paper, “A conceptual framework for trajectory-based medical analytics with IoT context”, by La presents a comprehensive case study of behaviour analytics in the personal healthcare domain, from data capturing using IoT devices to the context integration and data analysis in the backend service system for real-time decision making. We thank the Journal of Computer and System Sciences for supporting this special issue. We also thank all contributors and referees for their kind co-operation in helping this special issue.