巨量資料視覺化之研究

Abstract

目前針對 Big Data 的處理，多著重於快速處理三大面向的屬性資料(Volume、Velocity、Variety，簡稱 3 V )，及使用商業智慧分析，來發掘出能夠提高商業價值的策略。2012年三月美國白宮提出的「Big Data Initiative」想法，就是針對現今的極大量資料，研發出新的儲存、管理、分析等技術。而大資料的視覺化分析（Visual Analytics），更是美國政府部門重點研發項目。資料視覺化 (Data Visualizations) 是一個對於人腦最有效的輸入途徑，進而解讀大數據的模式，眾多呈現方式中，標籤雲或稱文字雲大概是最常見的一種資料視覺化方法，針對使用者輸入的文章或是網頁內容，分析其字詞出現的頻率，以字體大小表達出特定期間內最熱門的關鍵字，頻率越高字詞越大。 分析資料視覺化，它是一種相當有趣的技術，透過特殊的運算模式、演算法將各種數據、文字、資料轉換為各種圖表、影像，使得資料可以比較容易為人所理解，因此本論文將探討不同類型的資料呈現方式，研究人員會依據資料的特性，分為數值資料(numerical data)或數量資料(quantitative data)、類別資料 (categorical data)或定義資料(qualitative data)，藉此完成資料類型的分類，再來將資料對應到視覺屬性 (也就是資料編碼)，決定哪一種視覺屬性來表達資料類型是最有效率的，包含2D 與 3D 的圖形化資料呈現、即時性的報表產生工具、動態儀表板、資料視覺化動畫模擬等工具，論文最後將列出這些不同工具的優缺點評比。

Big data processing focuses on the three dimension (Volume, Velocity, and Variety, 3V for short), and the usage of business intelligence analysis is to enhance business value. The White House of United States announced the ideas of Big Data Initiative in March 2012 to study the technology of novel storage, management, and analysis in dealing with huge amounts of data, and visual analytics is just the focus of R & D projects in the U.S. government. Data visualization is one of the effective ways to understand for the human brain, and the tag cloud (word cloud) is the common form of data visualization among many presentations. These ways are to collect word and count their frequency of occurrence, and then show the most popular keywords by the font size within a specific period. The higher frequency will be represented by larger word. Data visualization is a very interesting technology that applies special computational mode and algorithm to transfer all kinds of data into various charts and images; that is, people can understand insights of data more easily. Therefore, this paper will discuss different types of data to render, and researchers can classify data by numerical data (quantitative data) and categorical data (qualitative data) according to their characteristics. By completing the classification of the data type, we map the data and its visible attributes (data coding process) to decide what kind of attributes to express the data type is the most efficient. These presentations include graphics rendering (2D and 3D), real-time reporting tools, dynamic dashboard, and data visualization tools such as simulation. Finally, we list the advantages and disadvantages of these different tools and their evaluation.