时空轨迹序列模式挖掘方法综述

在全球定位、移动通信技术迅速发展的背景下涌现出了海量的时空轨迹数据,这些数据是对移动对象在时空环境下的移动模式和行为特征的真实写照,蕴含了丰富的信息,这些信息对于城市规划、交通管理、服务推荐、位置预测等领域具有重要的应用价值,而时空轨迹数据在这些领域的应用通常需要通过对时空轨迹数据进行序列模式挖掘才能得以实现.时空轨迹...     

时空轨迹大数据模式挖掘研究进展

时空轨迹挖掘是数据挖掘领域的前沿研究课题,通过研究和开发时空轨迹挖掘技术,来发现隐藏在轨迹大数据中有价值的规律和知识以供决策支持。本文介绍了时空轨迹大数据模式挖掘与知识发现领域的研究进展;然后对时空轨迹模式挖掘技术产生的背景、应用领域和研究现状作了简介,并探讨了面向时空轨迹大数据模式挖掘的研究内容、系统架构以及关键技术,最后对时空轨迹频繁模式、伴随模式、聚集模式和异常模式的挖掘算法思想进行了阐述。     

基于网格索引的时空轨迹伴随模式挖掘算法

时空轨迹伴随模式是数据挖掘领域的一项重要研究内容。CMC(Coherent Moving Cluster)算法是一种经典的时空轨迹伴随模式挖掘算法,该算法引入了DBSCAN算法以挖掘出任意形状的簇。但是,DBSCAN聚类算法极耗时,导致CMC算法的时间效率较低。因此提出了一种基于网格索引的时空轨迹伴随模式挖掘算法MAP-G(Mining Adjoint Pattern of spatial-temporal trajectory based on the Grid index)。实验表明,MAP-G算法不仅比CMC算法具有更高的时间效率,而且能够过滤掉部分不正确的结果,因此结果也更加准确。     

轨迹数据的时空模式挖掘与管理决策研究综述

时空轨迹数据的获取变得越来越容易,轨迹数据刻画了移动对象的行为模式与活动规律,是对移动对象在时空环境下的移动模式和行为特征的真实写照,在城市规划、交通管理、服务推荐、位置预测等领域具有重要的应用价值.这些过程通常需要通过对时空轨迹数据进行模式挖掘才能得以实现.简述了轨迹数据挖掘的预处理和基本步骤,归纳了异常轨迹检测方法...     

基于时空周期模式挖掘的活动语义识别方法

传统的活动语义识别研究侧重从时空轨迹的空间信息中提取人类的活动语义,对时空轨迹数据的时间特性挖掘不足。本文兼顾时间和空间特征,提出了一种基于周期模式挖掘的活动语义识别方法。首先将分离出的活动轨迹数据通过空间距离进行密度聚类分成不同轨迹簇;然后,根据轨迹簇的时序特征挖掘个体对特定位置的访问周期,基于该访问周期,并结合在该位置的停留时间,及其附近兴趣点分布等特征构建分类模型,识别人类个体的活动语义。基于签到数据和仿真数据的实验结果表明,结合周期特征的活动语义识别方法相比没有加入周期特征的实验结果有效提升识别精度20%以上,在2个相同的签到数据集下,对比其他的识别方法提升精度10%以上。     

时空轨迹大数据分布式蜂群模式挖掘算法

针对时空轨迹大数据的蜂群模式挖掘需求,提出了一种高效的基于MapReduce的分布式蜂群模式挖掘算法。首先,提出了基于最大移动目标集的对象集闭合蜂群模式概念,并利用最小时间支集优化了串行挖掘算法;其次,提出了蜂群模式的并行化挖掘模型,利用蜂群模式时间域无关性,并行化了聚类与子时间域上的蜂群模式挖掘过程;第三,设计了一个基于MapReduce链式架构的分布式并行挖掘算法,通过四个阶段快速地实现了蜂群模式的并行挖掘;最后,在Hadoop平台上,使用真实交通轨迹大数据集对分布式算法的有效性和高效性进行了验证与分析。     

基于深度表示模型的移动模式挖掘

针对时空轨迹中位置顺序和时间对于理解用户移动模式的重要性,提出了一种新的用户轨迹深度表示模型。该模型考虑到时空轨迹的特点:1)不同的位置顺序表示不同的移动模式;2)轨迹有周期性并且在不同的时间段有变化。首先,将两个连续的位置点组合成位置序列;然后,将位置序列和对应的时间块组合成时间位置序列,作为描述轨迹特征的基本单位;最后,利用深度表示模型为每个序列训练特征向量。为了验证深度表示模型的有效性,设计实验将时间位置序列向量应用到用户移动模式发现中,并利用Gowalla签到数据集进行了实验评测。实验结果显示提出的模型能够发现"上班""购物"等明确的模式,而Word2Vec很难发现有意义的移动模式。     

基于时空图的移动对象聚集模式挖掘方法

移动对象聚集模式是指由移动对象参与的一组群体事件,通常用来预测交通系统中出现的异常现象.然而由于海量移动轨迹数据的产生,已有的研究方法难以准确、高效地挖掘特定的聚集模式.为此,提出一种基于时空图的移动对象聚集模式挖掘方法.该方法首先通过改进的空间聚类算法(DBScan)分析轨迹数据,从而获得移动对象聚类;然后,利用时空图模型代替单独存储轨迹数据的方式,用于实时观测移动对象聚类的时空变化特征.最后提出基于最大完全子图查找的聚集检索算法及其改进算法,用于查找满足时空约束的最大完全子图.基于真实大规模轨迹数据集上的实验结果表明,所提出的方法在移动对象聚集模式挖掘的准确性和高效性方面优于其他方法.     

面向位置预测的动态轨迹模式挖掘

针对海量的用户轨迹数据进行研究,提出一种动态分析移动对象轨迹模式、预测轨迹位置的方法（PRED）。首先使用改进的模式挖掘模型,提取轨迹频繁模式（简称T-模式）,然后提出DPTUpdate算法,设计蕴含时空信息的快捷数据结构--DPT（Dynamic Pattern Tree）,存储和查询移动物体的T-模式,并提出Prediction算法计算最佳匹配度,得到移动对象轨迹的预测位置。PRED方法可提供动态分析的能力,基于真实数据集进行对比实验,结果证明,平均准确率达到72%,平均覆盖率达到92.1%,与已有方法相比,其预测效果有显著提升。     

基于时空模式的轨迹数据聚类算法

针对轨迹聚类算法在相似性度量中多以空间特征为度量标准,缺少对时间特征的度量,提出了一种基于时空模式的轨迹数据聚类算法。该算法以划分再聚类框架为基础,首先利用曲线边缘检测方法提取轨迹特征点;然后根据轨迹特征点对轨迹进行子轨迹段划分;最后根据子轨迹段间时空相似性,采用基于密度的聚类算法进行聚类。实验结果表明,使用所提算法提取的轨迹特征点在保证特征点具有较好简约性的前提下较为准确地描述了轨迹结构,同时基于时空特征的相似性度量因同时兼顾了轨迹的空间与时间特征,得到了更好的聚类结果。     

Mining frequent trajectory patterns in spatial-temporal databases

In this paper, we propose an efficient graph-based mining (GBM) algorithm for mining the frequent trajectory patterns in a spatial-temporal database. The proposed method comprises two phases. First, we scan the database once to generate a mapping graph and trajectory information lists (TI-lists). Then, we traverse the mapping graph in a depth-first search manner to mine all frequent trajectory patterns in the database. By using the mapping graph and TI-lists, the GBM algorithm can localize support counting and pattern extension in a small number of TI-lists. Moreover, it utilizes the adjacency property to reduce the search space. Therefore, our proposed method can efficiently mine the frequent trajectory patterns in the database. The experimental results show that it outperforms the Apriori-based and PrefixSpan-based methods by more than one order of magnitude.     

Characterizing the state and processes of change in a dynamic forest environment using hierarchical spatio-temporal segmentation

Discrete changes in forest abundance, distribution, and productivity are readily detectable using a number of remotely sensed data sources; however, continuous changes such as growth and succession processes are more difficult to monitor. In this research we explore the potential of spectral trajectories generated from a 35-year (1973-2008) time-series of Landsat imagery to characterize change processes in a dynamic forest environment in northwestern Alberta, Canada. We propose a method of hierarchical spatio-temporal segmentation that enables the characterization of change processes that are spatially diffuse and temporally imprecise. Calibrated imagery from Landsat sensors are radiometrically normalized and two metrics derived from the Tasseled Cap Transformation components, greenness and brightness, are used to generate the Tasseled Cap Angle (TCA). The TCA is a measure of the proportion of vegetation to non-vegetation (the occupation state), and its derivative, the Process Indicator (PI), is a measure of change in this proportion through time. These indices condense information from the visible and near-infrared wavelengths, and facilitate lengthy time series analysis of forest landscape change using data from all Landsat sensors.A combination of the original TCA and its derivative sequence are input to a three level hierarchical segmentation process with the highest and lowest levels defining homogeneous objects at the initial and final date, and the intermediate level identifying trajectories with similar change processes. The development through time of the TCA and PI are described, and the spatial and temporal associations of processes are statistically assessed using the Moran's Index.A full range of change types were identified on the landscape, from stand replacing disturbances to more subtle growth and succession processes. Results indicate that the study area is in a constant state of change, and maintains a high average proportion of vegetation to non-vegetation. The amount of total landscape modified per decade increased from 18% and 14% in the 1970s and 1980s respectively, to more than 30% and 33% in the 1990s and 2000s. On average, the proportion of vegetation to non-vegetation was increasing prior to 1981, decreasing between 1981 and 1997, and increasing post-1997. There was a high degree of spatial autocorrelation amongst change processes, with a maximum Moran's I of 0.79 in 1973; landscape change became more spatially disperse and widespread after 1981. Temporal correlation of change processes was observed locally, with the period 1990-1995 having the most persistent change.     

改进的基于空间模式聚类的图像分割*

为了提高算法在含有一定噪声的图片中的分割功能,将一种隶属度函数计算方法加入到一种空间模式聚类算法当中,使原本已经充分挖掘图像空间信息的聚类算法,在含有一定噪声的图片上得到较好的分割效果。实验证明,修改后的算法提高了抗噪性能。     

Mining spatio-temporal patterns in object mobility databases

With the increasing use of wireless communication devices and the ability to track people and objects cheaply and easily, the amount of spatio-temporal data is growing substantially. Many of these applications cannot easily locate the exact position of objects, but they can determine the region in which each object is contained. Furthermore, the regions are fixed and may vary greatly in size. Examples include mobile/cell phone networks, RFID tag readers and satellite tracking. This demands techniques to mine such data. These techniques must also correct for the bias produced by different sized regions. We provide a comprehensive definition of Spatio-Temporal Association Rules (STARs) that describe how objects move between regions over time. We also present other patterns that are useful for mobility data; stationary regions and high traffic regions. The latter consists of sources, sinks and thoroughfares. These patterns describe important temporal characteristics of regions and we show that they can be considered as special STARs. We define spatial support to effectively deal with the problem of different sized regions. We provide an efficient algorithm—STAR-Miner—to find these patterns by exploiting several pruning properties. Responsible editors: Charles Perng and Tao Li.     

基于LCS的用户时空行为兴趣相似性计算方法

移动用户的位置轨迹中蕴含着用户的运动规律,行为模式等丰富的信息。重点关注用户轨迹中的停留区域,从轨迹间共同停留区域的时空重叠中挖掘用户行为兴趣的相似性,提出一种基于最长公共子序列的用户时空行为兴趣相似性计算方法。提取轨迹中的停留区域,利用最小包围盒技术描述轨迹中的停留区域,结合最长公共子序列算法提出一种基于最长公共子包围盒长度的用户相似性计算方法。实验收集60个志愿者6周的真实时空轨迹数据来评价该方法,实验结果表明该方法具有较高的准确率。