基于轨迹和POI数据的热点区域实时预测
发布时间：2018-09-08 06:51
【摘要】：目前,智能移动设备通常都带有利用了全球定位技术的位置传感器,可以准确的捕捉到移动设备的具体位置信息。随着位置获取和移动计算方面的技术的进步,产生了大量的数据,这些数据代表了各种不同种类的移动的物体如人,动物,车辆等的轨迹记录。并且这些位置信息可以通过众包的方式利用无线传输技术上传到服务器当中,形成大量的GPS轨迹数据。GPS轨迹大数据中包含了史无前例的信息,可以让我们更好的理解移动对象和地理位置,伴随产生了大量的基于位置的社交网络、智能交通系统和城市计算等应用。这些应用的流行,反过来又不断的促进新的系统性的轨迹大数据挖掘的研究。在这个良性的发展循环当中,GPS轨迹大数据的挖掘成为了一个热点的研究问题,吸引着包括计算机领域、社会学领域和地理信息领域的学者对此不懈研究。但是单纯从轨迹数据中挖掘人类活动语义信息是困难的。从更抽象的层次看,轨迹数据只是时空数据的一种,时空数据是带有空间坐标和时间戳的数据,具有时间属性和空间属性。能够准确预测时空数据变化情况对于地理位置推荐等城市计算相关应用具有重要意义。另一方面,城市中的兴趣点(Point of Interests,POI)等静态的地理空间信息也会对人类活动产生重要甚至是根本性的影响,因为人类是生活在一定地理环境中的,POI往往成为人类日常社会活动的参与场所。本文试图融合使用轨迹时空数据和地理静态信息两种数据,在复杂的城市环境中预测出不同区域的停留点数量变化情况。停留点也是一种时空数据,指出租车低速巡游或静止等待乘客的地点。停留点预测对于出租车载客,乘客乘车,地理位置的热度变化,乃至交通疏导,城市安全,城市建设规划等方面都具有重要参考作用。对停留点的预测往往是地理位置推荐系统中的一部分,但是现有的热点区域推荐或出租车载客点推荐等研究中往往在这些方面有所欠缺,一是使用数据源不够丰富,导致推荐准确率不够高,二是直接对于历史轨迹数据进行挖掘分析,计算复杂度高,难以拓展到更大地理范围,而且实时性难以保证。本文融合了交通轨迹数据和基于微博签到的POI数据,提出了自己的预测框架,框架首先根据基于POI的空间相似性找到相似区域,并根据相似区域预测出下个时间段目标区域的停留点变化情况。本文还基于Apache Storm构建了一个实时处理系统,模拟了整个实时分析与预测过程。实验结果表明,预测的地点的情况与实际情况相比具有较高的准确率,并且整个流处理系统也具有实时处理大数据的低延迟,高吞吐量的特性。
[Abstract]:At present, intelligent mobile devices are usually equipped with a position sensor using global positioning technology, which can accurately capture the specific location information of mobile devices. With the development of the technology of position acquisition and mobile computing, a lot of data are produced, which represent the track records of various kinds of moving objects such as human, animal, vehicle and so on. And these location information can be uploaded to the server by means of wireless transmission technology by crowdsourcing, forming a large amount of GPS trajectory data. Big data contains unprecedented information. It can give us a better understanding of mobile objects and geographical location, with a large number of location-based social networks, intelligent transportation systems and urban computing applications. The popularity of these applications, in turn, continues to promote the new systematic trajectory of big data mining research. In this benign cycle of development, the excavation of GPS trajectory big data has become a hot research issue, attracting scholars including computer field, sociology field and geographical information field to unremitting research on it. But it is difficult to mine the semantic information of human activities from track data. From a more abstract level, track data is only one kind of spatio-temporal data, and space-time data is data with spatial coordinates and timestamp, which has temporal and spatial attributes. It is very important to predict the change of spatiotemporal data accurately for the application of city calculation such as geographical location recommendation. On the other hand, static geospatial information such as (Point of Interests,POI) can have an important or even fundamental impact on human activities. Because people are living in a certain geographical environment, the POI often becomes the place of human daily social activities. This paper attempts to predict the change of the number of residence points in different regions in complex urban environment by using two kinds of data of locus spatiotemporal data and geographic static information. Stop points are also temporal and spatial data, indicating where taxis travel at low speeds or wait for passengers at rest. The prediction of stopping point plays an important reference role in the aspects of taxi passenger, passenger ride, heat change of geographical location, even traffic diversion, urban safety, urban construction planning and so on. The prediction of stopping point is often a part of the geographic location recommendation system, but the existing research on the recommendation of hot spot or taxi passenger spot is often lacking in these aspects. First, the use of data sources is not rich enough. As a result, the recommendation accuracy is not high enough, and the second is to mine and analyze the historical track data directly, which has high computational complexity and is difficult to extend to a larger geographical range, and the real-time performance is difficult to guarantee. This paper combines the traffic trajectory data with the POI data signed by Weibo, and proposes its own prediction framework. Firstly, the framework finds the similar region according to the spatial similarity based on POI. According to the similar region, the change of the residence point of the target area in the next time period is predicted. A real-time processing system based on Apache Storm is constructed, and the whole real-time analysis and prediction process is simulated. The experimental results show that the predicted location has a high accuracy compared with the actual situation, and the whole flow processing system also has the characteristics of real-time processing big data's low delay and high throughput.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13
本文编号：2229701
[Abstract]:At present, intelligent mobile devices are usually equipped with a position sensor using global positioning technology, which can accurately capture the specific location information of mobile devices. With the development of the technology of position acquisition and mobile computing, a lot of data are produced, which represent the track records of various kinds of moving objects such as human, animal, vehicle and so on. And these location information can be uploaded to the server by means of wireless transmission technology by crowdsourcing, forming a large amount of GPS trajectory data. Big data contains unprecedented information. It can give us a better understanding of mobile objects and geographical location, with a large number of location-based social networks, intelligent transportation systems and urban computing applications. The popularity of these applications, in turn, continues to promote the new systematic trajectory of big data mining research. In this benign cycle of development, the excavation of GPS trajectory big data has become a hot research issue, attracting scholars including computer field, sociology field and geographical information field to unremitting research on it. But it is difficult to mine the semantic information of human activities from track data. From a more abstract level, track data is only one kind of spatio-temporal data, and space-time data is data with spatial coordinates and timestamp, which has temporal and spatial attributes. It is very important to predict the change of spatiotemporal data accurately for the application of city calculation such as geographical location recommendation. On the other hand, static geospatial information such as (Point of Interests,POI) can have an important or even fundamental impact on human activities. Because people are living in a certain geographical environment, the POI often becomes the place of human daily social activities. This paper attempts to predict the change of the number of residence points in different regions in complex urban environment by using two kinds of data of locus spatiotemporal data and geographic static information. Stop points are also temporal and spatial data, indicating where taxis travel at low speeds or wait for passengers at rest. The prediction of stopping point plays an important reference role in the aspects of taxi passenger, passenger ride, heat change of geographical location, even traffic diversion, urban safety, urban construction planning and so on. The prediction of stopping point is often a part of the geographic location recommendation system, but the existing research on the recommendation of hot spot or taxi passenger spot is often lacking in these aspects. First, the use of data sources is not rich enough. As a result, the recommendation accuracy is not high enough, and the second is to mine and analyze the historical track data directly, which has high computational complexity and is difficult to extend to a larger geographical range, and the real-time performance is difficult to guarantee. This paper combines the traffic trajectory data with the POI data signed by Weibo, and proposes its own prediction framework. Firstly, the framework finds the similar region according to the spatial similarity based on POI. According to the similar region, the change of the residence point of the target area in the next time period is predicted. A real-time processing system based on Apache Storm is constructed, and the whole real-time analysis and prediction process is simulated. The experimental results show that the predicted location has a high accuracy compared with the actual situation, and the whole flow processing system also has the characteristics of real-time processing big data's low delay and high throughput.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13
【相似文献】
相关期刊论文 前4条
1 宋宣;;收购后阿里高德筹建POI库[J];信息与电脑(理论版);2014年01期
2 谢骥;;多系统合路平台(POI)测试方法探究[J];移动通信;2012年16期
3 赵卫锋;李清泉;李必军;;利用城市POI数据提取分层地标[J];遥感学报;2011年05期
4 何业军,田宇兴,卢军;POI在上海地铁CDMA接入系统中的应用[J];无线电工程;2002年04期
相关会议论文 前1条
1 王龙波;;基于GPS与实景影像的POI快速采集技术[A];全国测绘科技信息网中南分网第二十一次学术信息交流会论文集[C];2007年
相关硕士学位论文 前4条
1 周正;基于LBS的移动环境中POI推送方法研究[D];湖北工业大学;2015年
2 郑姗姗;移动通信中多系统合路平台（POI）及监控的研究[D];南京理工大学;2014年
3 龙涛;上海中心城区游客兴趣点（POI）时空分布特征研究[D];上海师范大学;2016年
4 孙瑞;基于轨迹和POI数据的热点区域实时预测[D];吉林大学;2017年
，本文编号：2229701
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2229701.html
