基于位置历史的个体生活模式挖掘

需积分: 9 178 浏览量更新于2024-09-09 收藏 1.1MB PDF 举报

在信息技术日益发展的今天，"Mining Individual Life Pattern Based on Location History"这篇文章探讨了随着GPS、GSM网络等位置获取技术的普及，人们如何方便地记录下自己的地理位置历史，从而转化为宝贵的空间时间数据。这一趋势促使研究者们关注如何从这些数据中挖掘出有价值的生活模式或规律。文章首先提出了个体生活模式（Individual Life Pattern, ILP）的概念，这是一种新颖的视角，旨在揭示个人普遍的生活习惯和日常活动的规律性。作者认识到，这种模式对于理解个人行为模式、提升用户体验以及进行个性化推荐等方面具有重要意义。为了系统地描述可以从位置历史中发现的生活规律，他们引入了生活模式的正常形式（Life Pattern Normal Form, LP-normal form），这是一种形式化的框架，用于规范性地表达可以从个人GPS数据中提取的各种类型的生活规则。接着，论文的核心部分是LP-Mine框架的提出。这个框架是一个有效的工具，它针对原始的个人GPS数据，设计了一系列算法和技术，包括数据预处理、模式识别和模式挖掘等步骤，以从海量的位置信息中高效地抽取个体生活模式。通过这种方式，研究人员可以识别出诸如常去的地点、出行习惯、活动时间模式等深层次的生活细节。该文还涉及到了数据隐私和安全问题，强调了在利用位置数据挖掘个体生活模式时必须遵循的伦理准则和法律框架，确保在尊重用户隐私的前提下进行研究。此外，文章还可能讨论了与其他数据源（如社交媒体、移动应用数据）相结合的可能性，以进一步增强生活模式的精准度和全面性。 "Mining Individual Life Pattern Based on Location History"是一篇在轨迹挖掘领域的重要研究论文，它不仅提供了理论基础，还提供了一套实用的方法论，对于理解个体行为、改进城市规划、提升服务个性化等方面具有深远的影响。通过深入分析和解读这篇论文，读者能够了解到如何运用现代信息技术从位置数据中洞察个体的生活模式，以及如何在技术和伦理之间找到平衡。

Mining Individual Life Pattern

Based on Location History

Yang Ye

#*1

, Yu Zheng

, Yukun Chen

, Jianhua Feng

, Xing Xie

Dept. Of Computer Science and Technology, Tsinghua University

Beijing, 100084, P.R. China

yey05@mails.tsinghua.edu.cn

fengjh@tsinghua.edu.cn

Microsoft Research Asia

4F, Sigma Building, No.49 Zhichun Road, Haidian District, Beijing 100190, P. R. China

{yuzheng,v-yukche, xing.xie}@microsoft.com

Abstract— The increasing pervasiveness of location-acquisition

technologies (GPS, GSM networks, etc.) enables people to

conveniently log their location history into spatial-temporal data,

thus giving rise to the necessity as well as opportunity to

discovery valuable knowledge from this type of data. In this

paper, we propose the novel notion of individual life pattern,

which captures individual’s general life style and regularity.

Concretely, we propose the life pattern normal form (the LP-

normal form) to formally describe which kind of life regularity

can be discovered from location history; then we propose the LP-

Mine framework to effectively retrieve life patterns from raw

individual GPS data. Our definition of life pattern focuses on

significant places of individual life and considers diverse

properties to combine the significant places. LP-Mine is

comprised of two phases: the modelling phase and the mining

phase. The modelling phase pre-processes GPS data into an

available format as the input of the mining phase. The mining

phase applies separate strategies to discover different types of

pattern. Finally, we conduct extensive experiments using GPS

data collected by volunteers in the real world to verify the

effectiveness of the framework.

I. INTRODUCTION

Nowadays, the development in location-acquisition

technologies and its embedding into people's daily life results

in a novel type of spatial-temporal data, which traces

individual location history and can be collected by the

wireless network infrastructures. For instance, when mobiles

phones are connected to GSM network, they left positioning

logs together with the timestamp of each log point. Likewise,

GPS-embedded portable devices can also record the latitude-

longitude position at every moment when exposed to a GPS

satellite. The increasing availability of individual location

history data bring us challenges as well as opportunities to

discover valuable knowledge from the raw data.

On this topic, some literatures aim at performing traditional

data mining tasks on spatial-temporal data, like classification

[1], clustering [2], pattern mining [3], [4] and outlier detection

[5]. In the meantime, some techniques have been proposed to

discovery higher level knowledge from individual GPS data

[15], including detecting significant locations of an individual,

predicting the movement destination [6], [11], recognize

individual mobility [14], etc.

However, the first class of research treats location history

data as general spatial-temporal trajectory; thereby loses some

of their inner properties. For instance, each log point in the

location history contains absolute time spot. However, either

Temporally Annotated Sequences in [7] or Trajectory Pattern

Mining in [3] discards this absolute time information and just

calculate the time interval between two points as time

annotation. While current works in the second class typically

mine first-level knowledge about position and mobility from

location history, like significant places, possible destinations,

attribute of mobility like stationary and walking.

Since location history data are individually generated,

given the close relationship between people's daily life and

geographic locations, we claim that one's general life style and

regularity can be discovered from his/her location history.

Resembling traditional definitions of frequent pattern in

transaction database [8], we term individual’s general life

style and regularity life pattern. In contrast to the first level

knowledge about position and mobility, life patterns represent

a higher level knowledge drawn from location history data.

The discovery of life pattern has a manifold of application

scenarios. To illustrate, life pattern reflects the regularity of

one individual, thus can help people better learn their way of

life; it can also be embedded into location recommender

system, context-based computing, precise advertising,

computer-aided schedule and route arrangement. For instance,

if Tom's life pattern about the time he goes to work is

discovered, his intelligent cell phone can automatically help

arrange the travelling route according to that day's traffic

condition. If Tom's general body-building time and place is

discovered, intelligent advertising system may choose that

moment to cast health products advertisement to his mobile

equipment. Also, computer-aided schedule and route

arrangement system can intelligently advice Tom to arrange

the time, route or position of new activities, given his general

life pattern. In the meantime, through collecting and analysing

life patterns of multiple individuals, a lot of statistical and

mining work can be done to discovery valuable knowledge

about social trends and generalities.

This paper is motivated by the increasing availability of

individual GPS data and the usefulness of life pattern. We aim

2009 Tenth International Conference on Mobile Data Management: Systems, Services and Middleware

DOI 10.1109/MDM.2009.11

2009 Tenth International Conference on Mobile Data Management: Systems, Services and Middleware

DOI 10.1109/MDM.2009.11

下载后可阅读完整内容，剩余9页未读，立即下载

wwloser

粉丝: 0
资源: 1

基于位置历史的个体生活模式挖掘

Mining frequent trajectory pattern based on vague space partition

matrix methods in data mining and pattern recognition

Machine Learning and Data Mining in Pattern Recognition 无水印pdf

分析一下Deep Feature Mining via Attention-based BiLSTM-GCN for Human Motor Imagery Recognition这篇文献，重点将模型实现过程阐述出来

pattern history table

Deep Feature Mining via Attention-based BiLSTM-GCN for Human Motor Imagery Recognition 这篇文献的创新点是什么

ECLAT: Frequent Pattern Mining with Vertical Data Format、

frequent pattern mining

最新资源