数据挖掘实战：机器学习与统计学解析

需积分: 32 186 浏览量更新于2024-07-24 收藏 6.94MB PDF 举报

"《Data Mining》是第三版的实用机器学习工具和技术的详细介绍，适合数据挖掘初学者，具有很高的权威性。" 数据挖掘是信息技术领域中的一个重要分支，它涉及到从大量数据中发现有价值的信息和知识的过程。这本书《Data Mining: Practical Machine Learning Tools and Techniques》深入浅出地讲解了这一领域的核心概念、技术和应用。作者Ian H. Witten、Eibe Frank和Mark A. Hall都是该领域的专家，他们的著作具有广泛的影响力。在数据挖掘中，机器学习是一个关键的组成部分，它涉及让计算机通过学习数据来改进其表现的能力。书中可能涵盖了监督学习、无监督学习和半监督学习等机器学习方法，如决策树、神经网络、支持向量机、朴素贝叶斯分类、聚类算法（如K-means）以及集成学习（如随机森林）等。统计学是数据挖掘的另一个基础，提供了理解数据分布、假设检验、回归分析等工具。书中的统计学部分可能会讨论如何使用这些工具来构建预测模型、进行关联规则学习以及异常检测。此外，书中可能还会涉及数据预处理，包括数据清洗、数据转换和特征选择，这些都是确保模型有效性的关键步骤。数据挖掘流程还包括数据探索、建模、评估和部署，这些都是实践中不可或缺的部分。数据挖掘的应用广泛，涵盖了商业智能、金融风险分析、医疗研究、社交媒体分析等多个领域。通过学习这本书，读者不仅可以理解基本的数据挖掘技术，还能学会如何在实际项目中应用这些知识，解决真实世界的问题。《Data Mining》第三版提供了一个全面的平台，让读者深入了解数据挖掘和机器学习的理论与实践，是入门者和专业人士的宝贵参考资料。书中可能包含丰富的实例和案例研究，帮助读者巩固理解，提升技能。同时，通过阅读此书，读者将能够更好地理解和运用统计学方法，提高在大数据时代处理和解读信息的能力。

Figure 1.1 Rules for the contact lens data. 12

Figure 1.2 Decision tree for the contact lens data. 13

Figure 1.3 Decision trees for the labor negotiations data. 18

Figure 2.1 A family tree and two ways of expressing the sister-of relation. 43

Figure 2.2 ARFF ﬁle for the weather data. 53

Figure 2.3 Multi-instance ARFF ﬁle for the weather data. 55

Figure 3.1 A linear regression function for the CPU performance data. 62

Figure 3.2 A linear decision boundary separating Iris setosas from Iris

versicolors. 63

Figure 3.3 Constructing a decision tree interactively. 66

Figure 3.4 Models for the CPU performance data. 68

Figure 3.5 Decision tree for a simple disjunction. 69

Figure 3.6 The exclusive-or problem. 70

Figure 3.7 Decision tree with a replicated subtree. 71

Figure 3.8 Rules for the iris data. 74

Figure 3.9 The shapes problem. 76

Figure 3.10 Different ways of partitioning the instance space. 80

Figure 3.11 Different ways of representing clusters. 82

Figure 4.1 Pseudocode for 1R. 86

Figure 4.2 Tree stumps for the weather data. 100

Figure 4.3 Expanded tree stumps for the weather data. 102

Figure 4.4 Decision tree for the weather data. 103

Figure 4.5 Tree stump for the ID code attribute. 105

Figure 4.6 Covering algorithm. 109

Figure 4.7 The instance space during operation of a covering algorithm. 110

Figure 4.8 Pseudocode for a basic rule learner. 114

Figure 4.9 Logistic regression. 127

Figure 4.10 The perceptron. 129

Figure 4.11 The Winnow algorithm. 130

Figure 4.12 A kD-tree for four training instances. 133

Figure 4.13 Using a kD-tree to ﬁnd the nearest neighbor of the star. 134

Figure 4.14 Ball tree for 16 training instances. 136

Figure 4.15 Ruling out an entire ball (gray) based on a target point

(star) and its current nearest neighbor. 137

Figure 4.16 A ball tree. 141

Figure 5.1 A hypothetical lift chart. 170

Figure 5.2 Analyzing the expected beneﬁt of a mailing campaign. 171

Figure 5.3 A sample ROC curve. 173

Figure 5.4 ROC curves for two learning schemes. 174

Figure 5.5 Effect of varying the probability threshold. 178

Figure 6.1 Example of subtree raising. 196

List of Figures

xvi List of Figures

Figure 6.2 Pruning the labor negotiations decision tree. 200

Figure 6.3 Algorithm for forming rules by incremental reduced-error

pruning. 207

Figure 6.4 RIPPER. 209

Figure 6.5 Algorithm for expanding examples into a partial tree. 210

Figure 6.6 Example of building a partial tree. 211

Figure 6.7 Rules with exceptions for the iris data. 213

Figure 6.8 Extended preﬁx trees for the weather data. 220

Figure 6.9 A maximum-margin hyperplane. 225

Figure 6.10 Support vector regression. 228

Figure 6.11 Example datasets and corresponding perceptrons. 233

Figure 6.12 Step versus sigmoid. 240

Figure 6.13 Gradient descent using the error function w

+ 1. 240

Figure 6.14 Multilayer perceptron with a hidden layer. 241

Figure 6.15 Hinge, squared, and 0 – 1 loss functions. 242

Figure 6.16 A boundary between two rectangular classes. 248

Figure 6.17 Pseudocode for model tree induction. 255

Figure 6.18 Model tree for a dataset with nominal attributes. 256

Figure 6.19 A simple Bayesian network for the weather data. 262

Figure 6.20 Another Bayesian network for the weather data. 264

Figure 6.21 The weather data. 270

Figure 6.22 Hierarchical clustering displays. 276

Figure 6.23 Clustering the weather data. 279

Figure 6.24 Hierarchical clusterings of the iris data. 281

Figure 6.25 A two-class mixture model. 285

Figure 6.26 DensiTree showing possible hierarchical clusterings of a given

dataset. 291

Figure 7.1 Attribute space for the weather dataset. 311

Figure 7.2 Discretizing the temperature attribute using the entropy

method. 318

Figure 7.3 The result of discretizing the temperature attribute. 318

Figure 7.4 Class distribution for a two-class, two-attribute problem. 321

Figure 7.5 Principal components transform of a dataset. 325

Figure 7.6 Number of international phone calls from Belgium, 1950–1973. 333

Figure 7.7 Overoptimistic probability estimation for a two-class problem. 344

Figure 8.1 Algorithm for bagging. 355

Figure 8.2 Algorithm for boosting. 359

Figure 8.3 Algorithm for additive logistic regression. 365

Figure 8.4 Simple option tree for the weather data. 366

Figure 8.5 Alternating decision tree for the weather data. 367

Figure 9.1 A tangled “web.” 391

Figure 11.1 The Explorer interface. 408

Figure 11.2 Weather data. 409

Figure 11.3 The Weka Explorer. 410

List of Figures xvii

Figure 11.4 Using J4.8. 411

Figure 11.5 Output from the J4.8 decision tree learner. 412

Figure 11.6 Visualizing the result of J4.8 on the iris dataset. 415

Figure 11.7 Generic Object Editor. 417

Figure 11.8 The SQLViewer tool. 418

Figure 11.9 Choosing a ﬁlter. 420

Figure 11.10 The weather data with two attributes removed. 422

Figure 11.11 Processing the CPU performance data with M5′. 423

Figure 11.12 Output from the M5′ program for numeric prediction. 425

Figure 11.13 Visualizing the errors. 426

Figure 11.14 Working on the segment-challenge data with the User

Classiﬁer. 428

Figure 11.15 Conﬁguring a metalearner for boosting decision stumps. 429

Figure 11.16 Output from the Apriori program for association rules. 430

Figure 11.17 Visualizing the iris dataset. 431

Figure 11.18 Using Weka’s metalearner for discretization. 443

Figure 11.19 Output of NaiveBayes on the weather data. 452

Figure 11.20 Visualizing a Bayesian network for the weather data

(nominal version). 454

Figure 11.21 Changing the parameters for J4.8. 455

Figure 11.22 Output of OneR on the labor negotiations data. 458

Figure 11.23 Output of PART for the labor negotiations data. 460

Figure 11.24 Output of SimpleLinearRegression for the CPU performance

data. 461

Figure 11.25 Output of SMO on the iris data. 463

Figure 11.26 Output of SMO with a nonlinear kernel on the iris data. 465

Figure 11.27 Output of Logistic on the iris data. 468

Figure 11.28 Using Weka’s neural-network graphical user interface. 470

Figure 11.29 Output of SimpleKMeans on the weather data. 481

Figure 11.30 Output of EM on the weather data. 482

Figure 11.31 Clusters formed by DBScan on the iris data. 484

Figure 11.32 OPTICS visualization for the iris data. 485

Figure 11.33 Attribute selection: specifying an evaluator and a search

method. 488

Figure 12.1 The Knowledge Flow interface. 496

Figure 12.2 Conﬁguring a data source. 497

Figure 12.3 Status area after executing the conﬁguration shown in

Figure 12.1. 497

Figure 12.4 Operations on the Knowledge Flow components. 500

Figure 12.5 A Knowledge Flow that operates incrementally. 503

Figure 13.1 An experiment. 506

Figure 13.2 Statistical test results for the experiment in Figure 13.1. 509

Figure 13.3 Setting up an experiment in advanced mode. 511

Figure 13.4 An experiment in clustering. 513

剩余664页未读，继续阅读

DataMining2013

粉丝: 0
资源: 1

数据挖掘实战：机器学习与统计学解析

数据仓库与DataMining技术结合的建模实践

DataMining入门：与统计分析的区别及与DataWarehousing的关系

Oracle Data Mining 11g Administrator指南

data mining

dataMining

datamining

DATAMINING

DataMining

Datamining

Oracle Data Mining 11g 概念与应用

最新资源