Link to Session Platform

Chairs(s):

Bandana Kar, Oak Ridge National Laboratory

; ,

Description:

The growth of information and communication technologies (ICT) has enabled citizen participation in scientific investigation (a.k.a. citizen science) and sharing of data and information via social media (e.g., Twitter) and social networking sites (e.g., Facebook). The advancements in Internet of Things (IoTs) and connected devices including drones and aerial robotics have enabled the use of social media generated big data to understand human dynamics, and their interaction with the built environments. Significant advancements have been made to collect and analyze these data for emergency response, risk communication, mobility studies among others.

The big data derived from citizen sensors tend to suffer from a myriad of uncertainties in terms of positional accuracy, context ambiguity, credibility, reliability, representativeness and completeness. Moreover, there are also serious concerns about data provenance and privacy. While there is no shortage in big data applications, the quality issue of these data remains an intellectual and practical challenge. A lack of data provenance for these data combined with unavailability of high-quality reference data appropriate to its enormous volume, heterogeneous structure in near real-time make it difficult to evaluate the quality of these data. Moreover, the notion of “ground truth” in social science research is subjected to the discourse of space-place dichotomy, the spatial and contextual randomness in human behaviors. The heterogeneous nature of these data in terms of data structure and content requires tremendous amount of processing at various stages of analytics before the data could be integrated with other geospatial datasets for decision-making purposes. Privacy awareness is of increasing importance to data management, dissemination and distribution in many research projects. Although aggregation, permutation or masking techniques can be used to protect data privacy without compromising the overall quality of data, its effectiveness depends on the degree of distribution heterogeneity of the geographic phenomenon. This session welcomes basic and empirical research that advances existing understanding and techniques to address the quality issue of big data generated from social media and its impact on applications pertaining to human dynamics, built environments and hazards. Possible topics may include but are not limited to:

• Quality issues in social media big data
• Challenges in collecting, processing and analyzing big data for real-time applications
• Big data quality and its impact in decision making
• Calibration and validation techniques/approaches in big data
• Data fusion of multi-source and/or heterogeneous datasets
• Big data analytics in hazards and built-environment
• Big data analytics in human movements and behaviors during disasters
• Geo-visualization techniques to analyze and visualize social media data
• Privacy and big data management
• Provenance and metadata generation
• Applications of machine-learning and computer vision in disaster research
• New methods to measure social media credibility of social media content and users
• Influential social media user detection

Non-Presenting Participants Agenda