Preface

17

Objective

17

Target Audience

18

Structure of This Book

19

1 An Introduction to Predictive Analytics

23

1.1 The Importance of Predictive Analysis

24

1.2 Predictive Analysis: Prescriptive and Exploratory

26

1.2.1 The Fundamental Idea

26

1.2.2 Prescriptive Analytics

28

1.2.3 Exploratory Analytics

29

1.2.4 Prescriptive vs. Exploratory Analytics

30

1.3 Preparing for a Successful Predictive Analysis Project

32

1.3.1 Stakeholders

32

1.3.2 Business Case: Objectives and Benefits

34

1.3.3 Requirements

35

1.3.4 Execution and Lifecycle Management

36

1.4 Industry Use Cases

37

1.5 Summary

39

2 What Is SAP Predictive Analytics?

41

2.1 Building Predictive Models with SAP Predictive Analytics

41

2.2 Automated Analytics and Expert Analytics

42

2.3 Mass Production of Predictive Models with the Predictive Factory

43

2.3.1 Result Production

43

2.3.2 Model Control

43

2.3.3 Model Retraining

44

2.3.4 Mass Production of Predictive Models

45

2.4 Data Preparation

45

2.5 Additional SAP Predictive Analytics Capabilities

47

2.6 Summary

48

3 Installing SAP Predictive Analytics

49

3.1 Recommended Deployments

50

3.2 Installing the SAP Predictive Analytics Server

51

3.2.1 Downloading the SAP Predictive Analytics Server

51

3.2.2 System Requirements

52

3.2.3 Installing the SAP Predictive Analytics Server

53

3.2.4 Post-Installation Steps

54

3.3 Installing the SAP Predictive Analytics Client

57

3.3.1 Downloading the SAP Predictive Analytics Client

58

3.3.2 System Requirements

58

3.3.3 Installation Steps

59

3.3.4 Checking the Installation

60

3.3.5 Starting the Client on Linux Operating Systems

60

3.4 Installing the Predictive Factory

61

3.4.1 Downloading the Predictive Factory

61

3.4.2 System Requirements

61

3.4.3 Installation Steps

62

3.4.4 Post-Installation Steps

63

3.5 Installing SAP Predictive Analytics Desktop

71

3.5.1 System Requirements

72

3.5.2 Installation Steps

72

3.5.3 Post-Installation Steps

73

3.6 SAP HANA Installation Steps

74

3.7 Summary

75

4 Planning a Predictive Analytics Project

77

4.1 Introduction to the CRISP-DM Methodology

78

4.2 Running a Project

81

4.2.1 Business Understanding

81

4.2.2 Data Understanding

85

4.2.3 Data Preparation

87

4.2.4 Modeling

93

4.2.5 Evaluation

96

4.2.6 Deployment

97

4.3 Summary

100

5 Predictive Factory

103

5.1 Predictive Factory: End-to-End Modeling

103

5.2 Creating a Project

105

5.3 External Executables

107

5.4 Variable Statistics

110

5.5 Summary

110

6 Automated Predictive Classification Models

111

6.1 Introducing Classification Models

112

6.1.1 The Classification Technique

112

6.1.2 Step-by-Step Classification Example

114

6.2 Creating an Automated Classification Model

115

6.2.1 Prerequisites

115

6.2.2 Creating the Data Connections and the Project

116

6.2.3 Creating the Model

117

6.3 Understanding and Improving an Automated Classification Model

123

6.3.1 Understanding an Automated Classification Model

123

6.3.2 Improving an Automated Classification Model

135

6.4 Applying an Automated Classification Model

139

6.4.1 Prerequisites

139

6.4.2 Applying a Classification Model

143

6.5 The Data Science behind Automated Predictive Classification Models

149

6.5.1 Foundations of Automated Analytics

150

6.5.2 Automated Data Preparation

160

6.5.3 Automated Data Encoding

164

6.6 Summary

172

7 Automated Predictive Regression Models

173

7.1 Introducing Regression Models

173

7.2 Creating an Automated Regression Model

174

7.3 Understanding and Improving an Automated Regression Model

179

7.3.1 Understanding an Automated Regression Model

179

7.3.2 Improving an Automated Regression Model

183

7.4 Applying an Automated Regression Model

184

7.4.1 Safely Applying an Automated Regression Model

184

7.4.2 Applying an Automated Regression Model

186

7.5 Summary

191

8 Automated Predictive Time Series Forecasting Models

193

8.1 Creating and Understanding Time Series Forecast Models

194

8.1.1 Creating and Training Models

195

8.1.2 Understanding Models

202

8.1.3 Saving Time Series Forecasts

206

8.1.4 Increasing Model Accuracy

210

8.2 Mass Producing Time Series Forecasts

214

8.3 Productizing the Forecast Model

221

8.4 The Data Science behind Automated Time Series Forecasting Models

227

8.4.1 Data Split

227

8.4.2 De-trending

228

8.4.3 De-cycling

229

8.4.4 Fluctuations

230

8.4.5 Smoothing

230

8.4.6 Model Quality

230

8.5 Summary

231

9 Massive Predictive Analytics

233

9.1 Deploying Predictive Models in Batch Mode

234

9.1.1 Deploying Times Series Forecasting Models

234

9.1.2 Deploying Classification/Regression Models

237

9.2 Model Quality and Deviation

241

9.2.1 Model Deviation Test Task Parameters

242

9.2.2 Model Deviation Test Task Outputs

242

9.3 Automatically Retraining Models

244

9.3.1 Defining a Model Retraining Task

244

9.3.2 Model Retraining Task Outputs

247

9.4 Scheduling and Combining Massive Tasks

250

9.4.1 Scheduling Tasks Independently

251

9.4.2 Event-Based Scheduling

253

9.5 Deploying Expert Analytics Models

253

9.6 Summary

256

10 Automated Analytics User Interface

259

10.1 When to Use Automated Analytics

259

10.2 Navigating the User Interface

260

10.3 Exploring the Automated Analytics Modules

264

10.3.1 Data Manager

264

10.3.2 Data Modeler

265

10.3.3 Toolkit

266

10.4 Summary

269

11 Automated Predictive Clustering Models

271

11.1 The Clustering Approach of Automated Analytics

272

11.2 Creating a Clustering Model

273

11.2.1 Starting the Clustering Module and Importing the Data

274

11.2.2 Analyzing the Dataset’s Content

279

11.2.3 Choosing the Variables and Setting the Model Properties

283

11.2.4 Generating the Model and Analyzing the Cluster Profiles

287

11.2.5 Applying the Model to a Dataset

293

11.2.6 Saving and Exporting the Model

295

11.3 Supervised and Unsupervised Clustering

297

11.4 The Data Science behind Automated Clustering Models

299

11.5 Summary

300

12 Social Network Analysis

301

12.1 Terminology of Social Network Analysis

302

12.2 Automated Functionalities of Social Network Analysis

303

12.2.1 Node Pairing

304

12.2.2 Communities and Roles Detection

304

12.2.3 Social Graph Comparison

305

12.2.4 Bipartite Graphs Derivation and Recommendations

305

12.2.5 Proximity

305

12.2.6 Path Analysis

306

12.3 Creating a Social Network Analysis Model

306

12.3.1 Starting the Module and Importing the Dataset

308

12.3.2 Defining the Graph Model to Build

310

12.3.3 Adding More Graphs

314

12.3.4 Setting Community, Mega-hub, and Node Pairing Detection

315

12.3.5 Providing Descriptions of Nodes

320

12.4 Navigating and Understanding the Social Network Analysis Output

323

12.4.1 Understanding Model Quality from the Model Overview

323

12.4.2 Navigating into the Social Network

328

12.4.3 Applying the Model

336

12.5 Colocation and Path Analysis Overview

340

12.5.1 Colocation Analysis

340

12.5.2 Frequent Path Analysis

344

12.6 Conclusion

346

13 Automated Predictive Recommendation Models

347

13.1 Introduction

348

13.1.1 Basic Concepts

348

13.1.2 Datasets

349

13.1.3 Recommended Approaches

351

13.2 Using the Social Network Analysis Module

353

13.2.1 Creating the Model

353

13.2.2 Understanding the Model

358

13.2.3 Applying the Model

365

13.3 Using the Recommendation Module

368

13.3.1 Creating the Model

368

13.3.2 Understanding the Model

369

13.3.3 Applying the Model

372

13.4 Using the Automated Predictive Library

374

13.5 Summary

375

14 Advanced Data Preparation Techniques with the Data Manager

377

14.1 Data Preparation for SAP Predictive Analytics

377

14.2 Building Datasets for SAP Predictive Analytics

379

14.2.1 Datasets

379

14.2.2 Methodology

380

14.3 Creating a Dataset using the Data Manager

381

14.3.1 Creating Data Manager Objects

383

14.3.2 Merging Tables

391

14.3.3 Defining Temporal Aggregates

394

14.3.4 Using the Formula Editor

398

14.4 Additional Functionalities

400

14.4.1 Visibility and Value

400

14.4.2 Domains

401

14.4.3 Documentation

402

14.4.4 Data Preview and Statistics

403

14.4.5 Generated SQL

406

14.4.6 Prompts

406

14.5 Using Data Manager Objects in the Modeling Phase

408

14.6 Managing Metadata

408

14.7 SQL Settings

411

14.8 Summary

412

15 Expert Analytics

415

15.1 When to Use Expert Analytics

415

15.2 Navigating the Expert Analytics Interface

416

15.3 Understanding a Typical Project Workflow

417

15.3.1 Selecting Data Source

418

15.3.2 Data Explorations

420

15.3.3 Graphical Components

422

15.3.4 Applying a Trained Model

424

15.3.5 Productionize

425

15.4 Creating an Expert Analytics Predictive Model

427

15.4.1 Load Data into SAP HANA

428

15.4.2 Create a Predictive Model

430

15.4.3 Model Deployment

436

15.5 Exploring the Available Algorithms

438

15.5.1 Connected to SAP HANA

438

15.5.2 Other Connectivity Types

440

15.6 Extending Functionality with R

442

15.6.1 Developing in an R Editor

442

15.6.2 Developing an R Function

444

15.6.3 Creating an R Extension

445

15.7 Summary

448

16 Integration into SAP and Third-Party Applications

451

16.1 Exporting Models as Third-Party Code

451

16.2 In-Database Integration

455

16.3 Scripting

456

16.3.1 Getting Started

456

16.3.2 KxShell Script

458

16.3.3 Executing Scripts

464

16.3.4 APIs

465

16.4 Summary

465

17 Hints, Tips, and Best Practices

467

17.1 Improving Predictive Model Quality

467

17.1.1 Data Quality

467

17.1.2 Data Description

468

17.1.3 Adding Predictors

468

17.1.4 Composite Variables

469

17.1.5 Adding Rows of Data

474

17.1.6 Reducing the Historical Period

474

17.1.7 Uniform Target Encoding

475

17.1.8 Segmented Modeling

475

17.1.9 Changing the Modeling Approach

475

17.2 Additional Resources

476

17.2.1 SAP Community and Newsletter

476

17.2.2 Product Documentation and Product Availability Matrix

476

17.2.3 Tutorials and Webinars

477

17.2.4 SAP Notes

478

17.3 Summary

478

18 Conclusion

479

18.1 Lessons Learned

479

18.2 The Future of SAP Predictive Analytics

479

18.3 Next Steps

480

Appendices

481

The Authors

481

Index

483