数据挖掘与聚类分析入门指南

需积分: 15 158 浏览量更新于2024-07-17 收藏 25.97MB PDF 举报

"Cluster Analysis and Data Mining: An Introduction" 是一本由 Ronald S. King 所著的专业书籍，专注于数据挖掘中的聚类分析。该书适用于行业专业人士的培训，也可作为聚类和分类课程的教学材料，无需预先具备聚类或数据挖掘的经验。书中重点介绍了非正式的聚类算法以及如何解释结果，利用图形方法和数据结构来评估聚类效果和探索数据。书中的例子和参考资料旨在让不同背景的读者都能理解。随书附带的光盘包含了练习答案、程序、数据集和图表等内容。标签 "Analytics" 指出这本书与数据分析相关，特别是聚类分析这一数据分析技术。书中的特色包括： 1. 强调在进行聚类分析时决策背后的逻辑。 2. 讨论统计学的应用，如Ward's方法（方差分析）、JAN（回归分析与相关性分析）和聚类验证（假设检验、拟合优度、蒙特卡洛模拟等）。 3. 单独章节专门探讨JAN和离散数据的聚类。 4. 光盘包含练习解答、程序、数据集和图表等资源。购买和使用本书需要遵守许可协议，该协议允许使用书中的内容，但不允许未经出版商书面同意就在互联网或其他网络上上传作品。复制或传播书中的文本、代码、模拟、图像等须遵守各自产品的许可条款，并需事先获得出版商或内容所有者的许可，才能复制或在网络上传播书中任何文本材料。出版商 Mercury Learning and Information（MLI）和参与创建、编写或生产随书光盘、算法、代码或计算机程序（“软件”）以及任何相关网站或软件的所有人员，对这些软件不提供任何形式的保证，也不承担因使用这些软件而产生的任何责任。

Preface • xv

The scheduling references for each of the chapters, in an applied statis-

tics class, could be as follows:

Chapters 1-4: after study of descriptive statistics.

Chapter 9: immediately following Chapters 1-3.

Chapter 6: after study of descriptive statistics.

Chapter 10: after studying the Empirical Rule and Chebychev’s Law.

Chapter 7: after studying probability.

Chapter 8: after study of hypothesis testing.

Chapter 5: after study of correlation, and both linear and multiple linear

regression.

Chapter 11: after study of statistical inference.

No previous experience or background in clustering is assumed. El-

ementary statistics plus a brief exposure to data structures are the prereq-

uisites. Informal algorithms for clustering data and interpreting results are

emphasized. In order to evaluate the results of clustering and to explore

data, graphical methods and data structures are used for representing data.

Throughout the text, examples and references are provided, in order to en-

able the material to be comprehensible for a diverse audience.

2 • Cluster Analysis and Data Mining

not needed. However, in the case where a single group decision is needed,

there must be a “meeting of the minds.”

One approach used is the Delphi Technique. This technique was designed

in the early 1950s by the Rand Corporation to predict future outcomes. It is

a group information gathering process to develop consensus opinion from a

panel of experts on a topic of interest. In the normal Delphi scenario, the pan-

el never meets face to face but interacts through questionnaires and feedback.

This noncontact approach alleviates the worry over such issues as individual

defensiveness or persuasiveness. However, opinions can be swayed due to a

participant observing the responses of the rest of the panel. Another problem

with the Delphi Technique is that the noncontact aspect is not feasible when,

for example, the panel is the graduate admissions committee at a university.

Cluster analysis is another technique that has been used with success

in the decision-making process. First, the investigator must determine the

answer to “What is a cluster?” The premise in cluster analysis is: given

a number of individuals, each of which is described by a set of numeri-

cal measures, devise a classification scheme for grouping the objects into a

number of classes such that the objects within classes are similar in some

respect and unlike those from other classes. These deduced classes are the

clusters. The number of classes and the characteristics of each class must

be determined from the data as discussed by Everett.

The key difference between cluster analysis and the Delphi Technique

is that cluster analysis is strictly an objective technique. Whereas individual

decisions can be swayed in an attempt to reach consensus in the Delphi

process, or a “happy medium” is reached which does not really portray the

feelings of the group as a whole. This is not the case in cluster analysis.

Clusters of individuals are reached using an objective mathematical func-

tion. One particular type of cluster analysis called Judgmental ANalysis

(JAN) takes the process one step further. Not only does it classify the panel

into similar groups based on a related regression equation, but it also allows

for these equations to be combined into a single policy equation. The JAN

technique has been in use since the 1960s. It has proven to be an effective

first step for methods of capturing and clustering the policies of judges.

Attempts at classification, that is sorting similar things into categories,

can be traced back to primitive humans. The ability to classify is a necessary

prerequisite for the development of language. Nouns, for example, are labels

Everitt, B. S. (1980). Cluster analysis (2nd ed.). New York: Halsted Press.

Introduction to Cluster Analysis • 3

used to classify a particular group of objects. Saying that a particular four-

legged animal is a “dog” allows us to put that animal into a category separate

from cats, sheep, and horses. In other words, it allows us to communicate.

The classification of people and animals is almost as old as language. The

early Hindus categorized humans into six types based on sex, physical, and

behavioral characteristics. The early Greeks and Romans used classification

to get a better understanding of the world around them. Galen, A.D. 129-199,

defined nine temperamental types that were assumed to be related to a per-

son’s susceptibility to various diseases and to individual differences in behav-

ior as discussed by Everitt.

Development of a method to categorize animals

into species was initiated by Aristotle. He started by dividing them into red

blooded (vertebrates) and those not having red blood (invertebrates). He

then subdivided the two groups again based on how their young were born.

Theophrastus continued Aristotle’s work, providing the groundwork for bio-

logical research for centuries. Eventually, new taxonomic systems were de-

veloped by such people as Linnaeus, Lindley, and Darwin. Classification

was not restricted to the biological sciences. In chemistry, Mendeleyev used

classification to develop the periodic tables, discussion by Everitt.

In the 1960s, two events led to an explosion of interest in cluster analy-

sis. The availability and spread of large, high-speed computers opened up

new possibilities for researchers. Additionally, the publication of Principles

of Numerical Taxonomy by Sokal and Sneath

covered the following three

important areas:

1. a number of different cluster analysis techniques

2. the use of computers in classification research

3. a radically empirical approach to biological taxonomy presented by

Blashfield and Aldenderfer

The need for cluster analysis arises in many fields of study. For example,

Anderberg

lists six areas where cluster analysis has been used successfully:

1. In the life sciences (biology, botany, zoology, etc.), the objects of

analysis are life forms such as plants, animals, and insects. The

Sokal, R. R., and Sneath, P. H. A. (1963). Principles of Numerical Taxonomy. W. H. Freeman.

Blashfield and Aldenderfer, M. S. (1978). The literature on cluster analysis. Multivariate

Behavioral Research, 13, 271-295.

Anderberg, M. R. (1973). Cluster analysis for applications. New York: Academic Press.

剩余332页未读，继续阅读

THESUMMERE

粉丝: 23
资源: 328

数据挖掘与聚类分析入门指南

Data Analytics with Hadoop: An Introduction for Data Scientists

Data.Mining.with.SPSS.Modeler.Theory.Exercises.and.Solutions.33192870

Data Clustering Introduction 1/2

The EM Algorithm and Extensions (2nd Edition)

数据分析与数据挖掘算法：K-means和层次聚类算法详解【英文版】

Cluster Analysis Evaluation: Silhouette Coefficient and Other Internal Metrics

【Advanced Chapter】Web Crawler Data Analysis and Visualization: Practical Implementation Using ...

[Practical Exercise] Data Storage and Analysis: Storing Scraped Data to Hadoop HDFS and Processing ...

【MATLAB Signal Preprocessing】: Data Cleaning and Noise Reduction Methods

MATLAB Reading JSON Data from TXT Files: A Powerful Tool for Data Exchange, Easily Parsing JSON ...

最新资源