Challenges and prospects of uncertainties in spatial big data analytics

Abstract

Knowledge extraction from spatial big data (SBD) with advanced analytics has become a major trend in research and industry. Meanwhile, the increasingly complex SBD and its analytics face proliferating challenges posed by uncertainties in them. Linked to various characteristics of SBD, the uncertainties emerge and propagate in each stage of SBD analytics. To avoid unreliable knowledge and losses resulting from the uncertainties and to ensure the value of authentic knowledge, this article proposes uncertainty-based SBD analytics. Uncertainty-based SBD analytics strive to understand, control, and alleviate uncertainties and their propagation in each stage of geographic knowledge extraction. Key topics involved in uncertainty-based SBD analytics include, for example, place-based heuristics for learning urban structure and place-based analytics on broader knowledge extraction tasks; dealing with the biases and inferencing the semantics in cell phone tracking data; quality assessment of unstructured spatial user-generated contents and the rectification of location shifts and time elapses between humans' activities and corresponding online contents they generate; and uncertainty handling in sophisticated black-box analytics with SBD such as deep learning. Challenges and the latest advances in each of these topics are presented, and further research for addressing these challenges is suggested in this article. 运用先进的分析从空间大数据(SBD)中获取知识,已成为研究和产业的主要趋势。于此同时,逐渐复杂化的SBD及其分析,因自身的不确定性而带来的挑战激增。不确定性与SBD的各种特徵相互连结,在SBD分析的各阶段中浮现并增生。为了避免不确定所导致的不可靠知识与损失,以及确保原初知识的价值,本文提出根据不确定性的SBD分析。根据不确定性的SBD分析,旨在理解、控制并减轻地理知识获取的各阶段中的不确定性及其增生。以不确定性为基础的SBD所涉及的主要主题,包含例如学习城市结构时以空间为基础的启发法,以及以地方为基础的更广泛的知识获取工作之分析;应对偏见并推断手机追踪数据中的语义学;非结构性空间使用者生产的内容的质量评估,并矫正人类活动及其生产的相应网路内容之间的区位变异和时间经过;以及在诸如深度学习的SBD之复杂黑箱分析中的不确定性处理。本文将呈现这些主题各自的挑战与最新的进展,并提出应对这些挑战的进一步研究之建议。