《数据可视化原理与实践第二版》深度解析

需积分: 9 17 浏览量更新于2024-07-17 收藏 59.33MB PDF 举报

《数据可视化原理与实践（第二版）》是Alexandru Telea所著的一本经典著作，于2007年由AK Peters Ltd首次发行。该书深入探讨了数据可视化的核心原则和实际应用，是数据科学、信息可视化以及科学可视化领域的重要参考书籍。在第二版中，作者将理论与实践紧密结合，提供了丰富的实例和深度解析，帮助读者理解和掌握如何有效地利用视觉元素来呈现和解释复杂的数据集。书中涵盖了广泛的议题，从基础概念如数据的多维表示到高级技术，如维度减少（DR）算法的应用。作者强调了数据可视化的关键原则，如清晰度、一致性、可读性和交互性，这些都是构建有效视觉化表达的基础。例如，章节11.5.7中提到的示例展示了如何通过结合科学可视化和信息可视化技术，使用二维投影展示高维数据，并通过Alpha融合的彩色纹理和图像基底技术来揭示假阳性或假阴性投影误差。书中还介绍了如何利用基于图像的着色技术，如图像阴影垫，来突出输入数据集中的重要主题，这有助于用户快速识别关键信息。同时，作者探讨了图形组合技术，如上文所述的捆绑图，用于强调所有对之间的错误，并结合几何和图像方法构造，以提供全面的视角。《数据可视化原理与实践（第二版）》不仅适合专业数据分析师、数据科学家，也对研究人员、数据工程师和设计师具有很高的参考价值。它不仅提供了理论框架，而且提供了实用的工具和技术，帮助读者在处理现代数据密集型环境中，提升信息的洞察力和决策质量。通过阅读这本书，读者能够提升数据可视化素养，更好地理解和沟通复杂的统计分析结果。

1.1. How Visualization Works 3

recommend the classics, also known as the Red Book [Shreiner et al. 03] and the

Blue Book [Shreiner 04].

We have decided to follow a toolkit-independent exposition of visualization

principles and techniques for several reasons. First, we believe that understand-

ing the main principles and working of data-visualization algorithms should not

be intermixed with the arduous process of learning the inevitably speciﬁc inter-

faces and assumptions of a software toolkit. Second, we do not assume that all

readers have the extensive programming knowledge typically required to mas-

ter the eﬃcient usage of some visualization software toolkits in use nowadays.

Finally, diﬀerent users have diﬀerent requirements and work in diﬀerent con-

texts, so the choice of a speciﬁc toolkit would inevitably limit the scope of the

presentation.

Last but not least, designing a complete visualization system involves many

subtle decisions. When designing a complex, real-world visualization system,

such decisions involve many types of constraints, such as performance, plat-

form (in)dependence, available programming languages and styles, user-interface

toolkits, input/output data format constraints, integration with third-party code,

and more. Although important for the success of a system design, such aspects

are not in the realm of data visualization but of software architecture, design,

and programming. All in all, we believe that presenting the ﬁeld of data visu-

alization in a manner as independently as possible from a toolkit choice makes

this book accessible to a broader, less specialized audience.

1.1 How Visualization Works

The purpose of visualization is to get insight, by means of interactive graph-

ics, into various aspects related to some process we are interested in, such as a

scientiﬁc simulation or some real-world process. There are many deﬁnitions of

visualization. Following Williams et al., visualization is “a cognitive process per-

formed by humans in forming a mental image of a domain space. In computer

and information science it is, more speciﬁcally, the visual representation of a

domain space using graphics, images, animated sequences, and sound augmenta-

tion to present the data, structure, and dynamic behavior of large, complex data

sets that represent systems, events, processes, objects, and concepts” [Williams

et al. 95].

In most applications, the path from the given process to the ﬁnal images is

quite complicated and involves a series of elaborate data-processing operations.

Ultimately, however, the visualization process produces one or several images

1.1. How Visualization Works 5

• Which are the minimum, maximum, or outliers of the values in a dataset,

and for which data points do they occur?

• What is the distribution of the values in a dataset?

• Do values of diﬀerent datasets exhibit correlations, and how much?

• How well do values in a dataset match a given model or pattern?

Quantitative vs. qualitative questions. At this point, readers may ask themselves

why we need visualization to answer concrete questions such as what is the

maximum value of a dataset, when a simple textual display would suﬃce. There

are, indeed, many cases when this is true and when simple tools such as a

text-based query-and-answer system work the best. However, there are also

many cases when visualization works better for answering concrete, quantitative

questions. In these cases, the answer to the question is not a single number

but typically a set of numbers. For example, while the question “what is the

value of a function f(x) at a given point x” can be answered by printing the

value, the question “what are the values of f (x) for all x in a given interval” is

best answered by plotting the graph of f(x) over that interval. Distributions,

correlations, and trends of a value set are also best understood when depicted

visually. Clearly, in some situations, task performance is better using visual

representations, whereas in others, text-based representations are more eﬃcient.

An early study by Larkin and Simon [Larkin and Simon 87] on the eﬀectiveness

of visual representations on human task performance outlined two ways in which

visual representations can outperform text-based ones:

• by substituting (rapid) perceptual inferences for diﬃcult logical inferences;

• by reducing the search for information required for task completion.

Although not exhaustive, their research gives some theoretical underpinning to

the intuitive appeal of using visualization to comprehend information.

In addition to quantitative questions, a large number of questions targeted by

visualization are of a qualitative nature, e.g., “given a medical scan of a patient,

are there any anomalies that may indicate clinical problems?” A typical answer

to this question would involve the discovery of patterns that have particular

characteristics in terms of shape, position, or data values, which a human expert

The title of their work, “Why a diagram is (sometimes) worth 10000 words,” is emblematic

for the well-known saying “A picture is worth a thousand words.”

6 1. Introduction

such as a medical doctor would classify as anomalous, based on his previous

clinical experience. In such cases, it is quite hard, if not impossible, to answer

the questions using fully automatic means, given the vague deﬁnition of the

question and the high variability of the input data. The decisional input of the

human expert, supported by interactive visualizations, is indispensable.

Exact vs. fuzzy questions. In scientiﬁc and engineering practice, questions typi-

cally range between extremely precise ones and very vague ones. Visualization

is useful for both types of questions. Although it can be argued that precise

questions can be answered with simple search-like queries that do not need any

visual support, it can also be argued that displaying the answer of such queries in

a visual manner makes them simpler to assimilate. Moreover, the visual support

can provide additional information that may not be explicitly requested by the

question but that can open broader perspectives on the studied problem. For ex-

ample, displaying the highest peak on a land map both by indicating the peak’s

height as a number and also its position on a 3D elevation visualization shows

not only the absolute height value but also how that value relates to the average

height on the map (how high it is compared to the overall landscape, how many

land points are at or close to the maximal height, and so on). This information,

although not directly requested, can help formulate subsequent questions, which

ultimately help the user to acquire a deeper understanding of the data at hand.

This is one of the meanings of the term “insight.”

Discover the unknown: In the second case, we may have no precise questions

about our process at hand. However, we are interested in (literally) looking at,

or examining, the information that this process provides. Why, one might ask,

should we look at some data if we have no concrete questions to ask about it?

There are several reasons for this. First, we may have had similar data in the

past that was interesting, helpful, or critical for a given application, and we may

want to ﬁnd out if the current data is of the same kind. Second, we may have the

feeling that, by examining a certain piece of data, we can acquire information

on the process that produced it that we simply cannot obtain by other means.

This role of visualization closely matches the perspective of a researcher who is

interested in studying a phenomenon in order to ﬁnd out novel facts and establish

unexpected correlations. This is a second meaning of the term “insight.”

The two types of questions are not separated in practice but rather serve two

types of complementary scenarios. In the ﬁrst scenario type, one has a num-

ber of precise questions to answer and is quite familiar with the type of data

and application. This scenario is best targeted by ﬁne-tuned visualizations an-

1.1. How Visualization Works 7

swering those precise questions. However, after answering such questions, one

may discover that some fundamental problem is still not solved, because the

questions were too precise. In this situation, switching to the more exploratory,

open-ended visualization scenario is a good continuation. This scenario follows

a typical bottom-up pattern, going from precise to more general queries. In the

second scenario type, the user receives a dataset that he is largely unfamiliar

with. In this case, the best approach is to start with some exploratory visualiza-

tion that presents a general overview of the data. After the general impression

is formed, the user can decide, based on particularities discovered in the general

view, which more speciﬁc, detailed visualization to use next, and thus which

more precise question to answer. This scenario follows a top-down pattern,

going from an overview to a detailed investigation. In the visualization ﬁeld,

this type of investigation is sometimes referred to as the visualization mantra

of “overview, zoom, and details-on-demand,” as coined by Shneiderman [Card

et al. 99, Bederson and Shneiderman 03].

Finally, an important drive for visualizing data is sheer curiosity. Visualiza-

tion methods often produce unexpected and beautiful imagery from many exotic

types of data. Such images are intriguing and interesting to the eye. When

examined in more detail, they can lead researchers to discover unexpected and

valuable things about the processes and datasets at hand.

Examples. Figure 1.2 shows four examples of visualizations, each targeting dif-

ferent types of questions, and illustrating diﬀerent visualization techniques. Fig-

ure 1.2(a) shows a visualization of average temperature and rainfall data over

Europe in July over the period 1960–1990. Temperature is shown by color (red

for large values, green for low values). Rainfall levels are indicated by the shaded

bumps, with the highest-looking bumps indicating maxima. The central bump

indicated by the arrow answers the question “where is the highest-rainfall re-

gion?” As this bump is green, we see that this region also has a mild tempera-

ture. Red, shallow bumps in the low part of the map show hot and dry regions.

This visualization is discussed next in Chapter 5, Section 5.4.1.

Figure 1.2(b) shows a visualization of ﬂuid ﬂow in a box. Fluid enters through

an inlet in the back, then ﬂows and swirls inside the box, and exits through an

outlet in the front. The drawn tubes indicate paths of diﬀerent ﬂuid particles

as they advance from the inlet towards the outlet. Tubes are colored by the

ﬂow speed (blue for low speeds, red for high speeds). To answer the question

“where are the fastest ﬂow regions?,” we can look for the red-colored tubes. This

image also allows us to discover a potentially unknown phenomenon: Close to

the outlet, the ﬂow splits into two distinct rotating layers, indicated by the two

剩余611页未读，继续阅读

demerzel

粉丝: 2
资源: 3

《数据可视化原理与实践第二版》深度解析

Augmented Reality Principles and Practice

QtDataVisualization

QtDataVisualization库使用cmake怎么构建

No module named PyQt5.QtDataVisualization

QtDataVisualization三维散点图

Data visualization解释

system.windows.forms.datavisualization.dll

qtdatavisualization教程

using system.windows.forms.datavisualization.charting;需要安装那个包

qt data visualization

最新资源