轨迹大数据异常检测:研究进展及系统框架

发布时间：2018-04-10 00:06

本文选题：异常检测　切入点：轨迹大数据　出处：《软件学报》2017年01期

【摘要】：定位技术与普适计算的蓬勃发展催生了轨迹大数据,轨迹大数据表现为定位设备所产生的大规模高速数据流.及时、有效地对以数据流形式出现的轨迹大数据进行分析处理,可以发现隐含在轨迹数据中的异常现象,从而服务于城市规划、交通管理、安全管控等应用.受限于轨迹大数据固有的不确定性、无限性、时变进化性、稀疏性和偏态分布性等特征,传统的异常检测技术不能直接应用于轨迹大数据的异常检测.由于静态轨迹数据集的异常检测方法通常假定数据分布先验已知,忽视了轨迹数据的时间特征,也不能评测轨迹大数据中动态演化的异常行为.面对轨迹大数据低劣的数据质量和快速的数据更新,需要利用有限的系统资源处理因时变带来的概念漂移,实时地检测多样化的轨迹异常,分析轨迹异常间的因果联系,继而识别更大时空区域内进化的、关联的轨迹异常,这是轨迹大数据异常检测的核心研究内容.此外,融合与位置服务应用相关的多源异质数据,剖析异常轨迹的起因以及其隐含的异常事件,也是轨迹大数据异常检测当下亟待研究的问题.为解决上述问题,对轨迹异常检测技术的研究成果进行了分类总结.针对现有轨迹异常检测方法的局限性,提出了轨迹大数据异常检测的系统架构.最后,在面向轨迹流的在线异常检测、轨迹异常的演化分析、轨迹异常检测系统的基准评测、异常检测结果语义分析的数据融合以及轨迹异常检测的可视化技术等方面探讨了今后的研究工作.
[Abstract]:The rapid development of positioning technology and pervasive computing has given birth to big data, the locus big data, which is a large scale high-speed data stream produced by positioning equipment.By analyzing and dealing with the trajectory big data in the form of data flow in time and effectively, we can find the abnormal phenomenon hidden in the track data, which can serve the application of urban planning, traffic management, safety control and so on.Limited by the inherent uncertainty, infinity, time-varying evolution, sparsity and skewness of trajectory big data, the traditional anomaly detection techniques can not be directly applied to the trajectory big data anomaly detection.Because the anomaly detection method of static track data set usually assumes that the data distribution is known a priori, it neglects the time characteristic of track data, and can not evaluate the abnormal behavior of dynamic evolution in trajectory big data.In the face of big data's poor data quality and rapid data updating, it is necessary to use limited system resources to deal with the concept drift caused by time-varying, to detect various trajectory anomalies in real time, and to analyze the causal relationship between trajectory anomalies.Then we can identify the related trajectory anomalies in the larger space-time region, which is the core research content of trajectory big data anomaly detection.In addition, the fusion of multi-source heterogeneous data related to location service application, the analysis of the cause of abnormal trajectory and the hidden abnormal events are also the problems that need to be studied urgently in the detection of track big data anomaly.In order to solve the above problems, the research results of trajectory anomaly detection are classified and summarized.Aiming at the limitation of the existing trajectory anomaly detection methods, the system architecture of trajectory big data anomaly detection is proposed.Finally, on line anomaly detection for trajectory flow, evolution analysis of trajectory anomaly, benchmark evaluation of trajectory anomaly detection system,The data fusion of semantic analysis of anomaly detection results and the visualization technology of track anomaly detection are discussed in this paper.
【作者单位】：华东师范大学数据科学与工程学院;西华师范大学计算机学院;
【基金】：国家自然科学基金(61370101,U1501252,U1401256) 上海市教委创新计划(14ZZ045) 西华师范大学国家级项目培育专项(16C005)~~
【分类号】：TP311.13

【相似文献】