應用於大規模人群移動之視覺化與分析及建模技術之研究---總計畫暨子計畫一：大量動態資料視覺化中時變特徵之分析與表述

Abstract

隨著數位感測裝置的普及，加上網際網路與雲端科技的運用，數位裝置造就了大量 的資訊流。這些資料不只容量上非常巨大，内容與形式也相當多元，產生的速度快。如 何運用這龐大的資料流找出有用信息將對現代企業決策分析、災難管理、國家安全產生 重要影響。傳統的資料探勘與統計分析可以由演算法分析出巨量資料中所含的有用的隱 含訊息，但資料過於龐大與複雜且是時變資料流，傳統的方法將力有未逮或耗時過多、 或無法精確的分析出有用的訊息。

本整合型計晝針對巨量人群移動資料難以全自動化方法分析的問題，從資料互動視 覺化(Visualization)的角度出發，希望以視覺化及互動方式協助使用者作特徵與樣板分 析、移動動態分析、與模擬建構與驗證，並發展視覺化技術將巨量資料中有用的信息呈 現給使用者。本整合型計晝共有下列四個子計晝：

1.子計晝一『大量動態資料視覺化中時變特徵分析與表述』結合自動化與互動視覺化 的方法，以用於樣板分析、資訊標記與動態臆測；並將人群移動路線的地理拓樸結 合入分析資訊中。

2.子計晝二『結合互動式視覺化技術之交通與人群動態分析』將以人群在道路或大眾 運輸系統移動之視覺化動態分析為主，並探討人群移動與車輛交通的互動關係。針 對現今車輛交通及人群移動資料的大資料特性，利用互動式可視化技術來協助專家 有效呈現及分析交通及人群移動資料。

3.子計晝三『基於資料及模型混合驅動之人群與交通電腦模擬』將建立人群與交通電 腦動晝模擬系統，參考現有資料中擷取出的車輛軌跡、速度和人群等資訊，使用三 維動晝和視覺化技術呈現。

4. 子計晝四『人群移動網路視覺化』將透過視覺方法來分析、内插、及歸納人群移動的 資料。透過人類對視覺資料強大的理解能力，經由不同的角度來觀察及探索資料，以 期發現全自動的資料探勘技術所無法發現的珍貴訊息。

總計晝將負責使用者需求收集與資料庫服務架設，發展巨量資料視覺化顯像架構與 介面，並作使用者調查與回饋驗證分析。

Digital sensing devices have become very popular nowadays, which, through internet and cloud computing, produces large-scale data streams. The data stream is generally in big volume, has unstructured format and is of great diversity, and is being produced at a very fast rate. How to explore these big streaming data and find out useful information is crucial in business decision making, disaster management, cyber and national security. Traditional automatic analysis using data mining and statistical approaches is facing difficulties to deal with such a big data that is being generated at such fast rates.

The proposed integrated project focuses on large-scaled human mobility data, aiming to provide visual means to direct or assist feature/pattern analysis, mobility analysis, simulation modeling, and provide effective visualization viewing to users for summarizing information and revealing insight information from the big data. The proposed integrated project has the following four subprojects.

1. Subproject 1 『Time-Varying Feature Analysis and Representation in Large-Scale

Mobility Data Visualization』will combine automatic extraction and visualization with user feedback. The strategy will be used for pattern analysis, data labeling, and mobility prediction. Path topology will be integrated to improve the mobility analysis.

2. Subproject 2『Traffic and Human Mobility Analysis with Interactive Visualization』will

develop visual analytic techniques for analyzing human mobility that mainly travels via roads or public transportation systems. In addition, visual analytic techniques will be developed for studying the interaction between crowd movement and vehicle traffic.

3. Subproject 3『A Hybrid Approach for Crowd and Traffic Simulation』will build a

simulation system of crowd and traffic. The system can simulate and visualize the motion of the traffic and crowd in a complex environment, e.g. road map in a city. The visualization techniques can be adopted for extracting the important events and features from the simulated results.

4. Subproject 4『Human Mobility Visualization』will develop visual means for analyzing,

interpreting, and summarizing the human mobility data, which are typically dynamic and unstructured. The techniques developed will allow users to observe and explore data from different viewpoints through an interactive interface and discover important messages that are unavailable from fully automatic methods.

The main project will be responsible for data gathering, study of user’s requirement, providing storage system and service for large-scale human mobility data analysis, developing a general framework for large data visualization, and conducting a user-study and feedback evaluation.