融合复合特征的移动轨迹预测方法的研究与实现

发布时间：2018-01-10 09:26

本文关键词：融合复合特征的移动轨迹预测方法的研究与实现　出处：《西安电子科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着宽带无线接入技术和移动终端技术的飞速发展,人们迫切希望能够随时随地在移动过程中享受互联网的信息和服务,移动互联网应运而生并迅猛发展。各种基于位置服务的设备都要求提供移动设备的准确位置。目前,虽然定位系统的可靠性和准确性有所提高,但是由于GPS系统、移动设备、无线网络的局限性,定位系统有时难以精确追踪物体,需要可靠的方法预测移动对象的将来位置。本文提出了两种移动对象轨迹预测的方法:融合语义特征的移动对象轨迹预测方法SG和融合动态环境的移动对象概率路径预测方法P3D。融合语义特征的移动用户轨迹预测方法SG首先将用户的地理轨迹转化成包含语义行为的轨迹,挖掘出语义模式集,同时在语义轨迹中分析移动用户的公共行为,将具有相似语义行为的用户进行聚类,挖掘出每个聚类的地理模式集。基于挖掘到的用户个体语义模式集和相似用户地理模式集,构造用来索引和局部匹配的模式树STP-Tree和SLP-Tree。通过对STP-Tree和SLP-Tree的索引和局部匹配,引入一个加权函数对给定移动用户的最近运动进行预测。融合动态环境的移动对象概率路径预测方法P3D通过动态选择轨迹来动态建立预测模型。P3D的优点:1)要被预测的目标轨迹在模型建立之前被选择,可以针对和目标轨迹相关的轨迹建立模型。2)不像惰性学习方法,P3D基于少量选定的参考轨迹,用精确的学习方法在可接受的时延内得到准确的预测模型。3)如果预测运动和准确的结果不匹配,P3D可以通过动态重构建新模型进行持续自校正。在大量真实和人工轨迹数据集上的实验结果表明:本文SG方法的准确性和性能较传统方法都有显著提高。P3D通过自校正持续预测,可以动态调整预测结果,相比没有自校正持续预测方法,性能有了明显的提升。
[Abstract]:With the rapid development of broadband wireless access technology and mobile terminal technology, people are eager to enjoy the information and services of Internet anytime and anywhere. Mobile Internet arises at the historic moment and develops rapidly. All kinds of location-based devices are required to provide the exact location of mobile devices. At present, the reliability and accuracy of positioning system have been improved. However, due to the limitations of GPS systems, mobile devices and wireless networks, positioning systems sometimes find it difficult to track objects accurately. A reliable method is needed to predict the future position of moving objects. In this paper, two methods of trajectory prediction for moving objects are proposed:. The mobile object trajectory prediction method SG which fuses semantic features and the mobile object probabilistic path prediction method P3D. the mobile user trajectory prediction method based on semantic feature is firstly applied to user's geography by combining semantic features with SG and moving object probabilistic path prediction method based on dynamic environment. The locus is transformed into a locus containing semantic behavior. At the same time, the common behaviors of mobile users are analyzed in the semantic locus, and the users with similar semantic behaviors are clustered. Mining the geographical pattern set of each cluster. Based on the user individual semantic pattern set and similar user geographical pattern set. Construct pattern trees STP-Tree and SLP-Tree. for indexing and local matching by indexing and local matching of STP-Tree and SLP-Tree. A weighted function is introduced to predict the recent movement of a given mobile user. A method for predicting the probabilistic path of moving objects in a dynamic environment P3D dynamically establishes the prediction model by dynamically selecting the trajectory. P3D. Point:. 1). The target trajectory to be predicted is selected before the model is built. A model. 2) can be built for the trajectory associated with the target trajectory.) unlike the lazy learning method, P3D is based on a small number of selected reference trajectories. An accurate prediction model. 3 is obtained by using an accurate learning method within an acceptable time delay) if the predicted motion and the exact result do not match. P3D can be continuously self-corrected by dynamic reconstruction of new models. Experimental results on a large number of real and artificial trajectory data sets show that:. Compared with traditional methods, the accuracy and performance of SG method in this paper are significantly improved. P3D through self-tuning continuous prediction. The prediction results can be adjusted dynamically. Compared with no self-tuning continuous prediction method, the performance has been improved significantly.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN925.93

【共引文献】