Machine Learning for Business Analytics: Concepts, Techniques and Applications with Jmp Pro

ISBN-13: 9781119903833 / Angielski / Twarda / 2023 / 600 str.

Nitin R. Patel

Spis treści

Foreword xixPreface xxAcknowledgments xxiiiPart I Preliminaries1 Introduction 31.1 What Is Business Analytics? 31.2 What Is Machine Learning? 51.3 Machine Learning, AI, and Related Terms 5Statistical Modeling vs. Machine Learning 61.4 Big Data 61.5 Data Science 71.6 Why Are There So Many Different Methods? 81.7 Terminology and Notation 81.8 Road Maps to This Book 10Order of Topics 122 Overview of the Machine Learning Process 172.1 Introduction 172.2 Core Ideas in Machine Learning 18Classification 18Prediction 18Association Rules and Recommendation Systems 18Predictive Analytics 19Data Reduction and Dimension Reduction 19Data Exploration and Visualization 19Supervised and Unsupervised Learning 192.3 The Steps in A Machine Learning Project 212.4 Preliminary Steps 22Organization of Data 22Sampling from a Database 22Oversampling Rare Events in Classification Tasks 23Preprocessing and Cleaning the Data 232.5 Predictive Power and Overfitting 29Overfitting 29Creation and Use of Data Partitions 312.6 Building a Predictive Model with JMP Pro 34Predicting Home Values in a Boston Neighborhood 34Modeling Process 362.7 Using JMP Pro for Machine Learning 422.8 Automating Machine Learning Solutions 43Predicting Power Generator Failure 44Uber's Michelangelo 452.9 Ethical Practice in Machine Learning 47Machine Learning Software: The State of the Market by HerbEdelstein 47Problems 52Part II Data Exploration and Dimension Reduction3 Data Visualization 593.1 Introduction 593.2 Data Examples 61Example 1: Boston Housing Data 61Example 2: Ridership on Amtrak Trains 623.3 Basic Charts: Bar Charts, Line Graphs, and Scatter Plots 62Distribution Plots: Boxplots and Histograms 64Heatmaps 673.4 Multidimensional Visualization 70Adding Variables: Color, Hue, Size, Shape, Multiple Panels,Animation 70Manipulations: Rescaling, Aggregation and Hierarchies, Zooming,Filtering 73Reference: Trend Line and Labels 77Scaling Up: Large Datasets 79Multivariate Plot: Parallel Coordinates Plot 80Interactive Visualization 803.5 Specialized Visualizations 82Visualizing Networked Data 82Visualizing Hierarchical Data: More on Treemaps 83Visualizing Geographical Data: Maps 843.6 Summary: Major Visualizations and Operations, According toMachine Learning Goal 87Prediction 87Classification 87Time Series Forecasting 87Unsupervised Learning 88Problems 894 Dimension Reduction 914.1 Introduction 914.2 Curse of Dimensionality 924.3 Practical Considerations 92Problems 112Part III Performance Evaluation5 Evaluating Predictive Performance 1175.1 Introduction 1185.2 Evaluating Predictive Performance 118Problems 142Part IV Prediction and Classification Methods6 Multiple Linear Regression 1476.1 Introduction 1476.2 Explanatory vs. Predictive Modeling 1486.3 Estimating the Regression Equation and Prediction 149Example: Predicting the Price of Used Toyota CorollaAutomobiles 1506.4 Variable Selection in Linear Regression 155Reducing the Number of Predictors 155How to Reduce the Number of Predictors 156Manual Variable Selection 156Automated Variable Selection 157Regularization (Shriknage Models) 164Problems 1707 k-Nearest Neighbors (k-NN) 1757.1 The k-NN Classifier (Categorical Outcome) 175Problems 1868 The Naive Bayes Classifier 1898.1 Introduction 189Threshold Probability Method 190Conditional Probability 190Problems 2039 Classification and Regression Trees 2059.1 Introduction 206Tree Structure 206Decision Rules 207Classifying a New Record 2079.2 Classification Trees 207Recursive Partitioning 207Example 1: Riding Mowers 208Categorical Predictors 210Standardization 2109.3 Growing a Tree for Riding Mowers Example 210Choice of First Split 211Choice of Second Split 212Final Tree 212Using a Tree to Classify New Records 2139.4 Evaluating the Performance of a Classification Tree 215Example 2: Acceptance of Personal Loan 2159.5 Avoiding Overfitting 219Stopping Tree Growth: CHAID 220Growing a Full Tree and Pruning It Back 220How JMP Pro Limits Tree Size 2219.6 Classification Rules from Trees 2229.7 Classification Trees for More Than Two Classes 2249.8 Regression Trees 224Prediction 224Evaluating Performance 2259.9 Advantages and Weaknesses of a Single Tree 2279.10 Improving Prediction: Random Forests and Boosted Trees 229Random Forests 229Boosted Trees 230Problems 23310 Logistic Regression 23710.1 Introduction 23710.2 The Logistic Regression Model 23910.3 Example: Acceptance of Personal Loan 240Model with a Single Predictor 241Estimating the Logistic Model from Data: Multiple Predictors 243Interpreting Results in Terms of Odds (for a Profiling Goal) 24610.4 Evaluating Classification Performance 24710.5 Variable Selection 24910.6 Logistic Regression for Multi-class Classification 250Logistic Regression for Nominal Classes 250Logistic Regression for Ordinal Classes 251Example: Accident Data 25210.7 Example of Complete Analysis: Predicting Delayed Flights 253Data Preprocessing 255Model Fitting, Estimation, and Interpretation---A Simple Model 256Model Fitting, Estimation and Interpretation---The Full Model 257Model Performance 257Problems 26411 Neural Nets 26711.1 Introduction 26711.2 Concept and Structure of a Neural Network 26811.3 Fitting a Network to Data 269Example 1: Tiny Dataset 269Computing Output of Nodes 269Preprocessing the Data 272Training the Model 273Using the Output for Prediction and Classification 279Example 2: Classifying Accident Severity 279Avoiding Overfitting 28111.4 User Input in JMP Pro 28211.5 Exploring the Relationship Between Predictors and Outcome 28411.6 Deep Learning 285Convolutional Neural Networks (CNNs) 285Local Feature Map 287A Hierarchy of Features 287The Learning Process 287Unsupervised Learning 288Conclusion 28911.7 Advantages and Weaknesses of Neural Networks 289Problems 29012 Discriminant Analysis 29312.1 Introduction 293Example 1: Riding Mowers 294Example 2: Personal Loan Acceptance 29412.2 Distance of an Observation from a Class 29512.3 From Distances to Propensities and Classifications 29712.4 Classification Performance of Discriminant Analysis 30012.5 Prior Probabilities 30112.6 Classifying More Than Two Classes 303Example 3: Medical Dispatch to Accident Scenes 30312.7 Advantages and Weaknesses 306Problems 30713 Generating, Comparing, and Combining Multiple Models 31113.1 Ensembles 311Why Ensembles Can Improve Predictive Power 312Simple Averaging or Voting 313Bagging 314Boosting 315Stacking 316Advantages and Weaknesses of Ensembles 31713.2 Automated Machine Learning (AutoML) 317AutoML: Explore and Clean Data 317AutoML: Determine Machine Learning Task 318AutoML: Choose Features and Machine Learning Methods 318AutoML: Evaluate Model Performance 320AutoML: Model Deployment 321Advantages and Weaknesses of Automated Machine Learning 32213.3 Summary 322Problems 323Part V Intervention and User Feedback14 Interventions: Experiments, Uplift Models, and Reinforcement Learning 32714.1 Introduction 32714.2 A/B Testing 328Example: Testing a New Feature in a Photo Sharing App 329The Statistical Test for Comparing Two Groups (T-Test) 329Multiple Treatment Groups: A/B/n Tests 333Multiple A/B Tests and the Danger of Multiple Testing 33314.3 Uplift (Persuasion) Modeling 333Getting the Data 334A Simple Model 336Modeling Individual Uplift 336Creating Uplift Models in JMP Pro 337Using the Results of an Uplift Model 33814.4 Reinforcement Learning 340Explore-Exploit: Multi-armed Bandits 340Markov Decision Process (MDP) 34114.5 Summary 344Problems 345Part VI Mining Relationships Among Records15 Association Rules and Collaborative Filtering 34915.1 Association Rules 349Discovering Association Rules in Transaction Databases 350Example 1: Synthetic Data on Purchases of Phone Faceplates 350Data Format 350Generating Candidate Rules 352The Apriori Algorithm 353Selecting Strong Rules 353The Process of Rule Selection 356Interpreting the Results 358Rules and Chance 359Example 2: Rules for Similar Book Purchases 36115.2 Collaborative Filtering 362Data Type and Format 363Example 3: Netflix Prize Contest 363User-Based Collaborative Filtering: "People Like You" 365Item-Based Collaborative Filtering 366Evaluating Performance 367Advantages and Weaknesses of Collaborative Filtering 368Collaborative Filtering vs. Association Rules 36915.3 Summary 370Problems 37216 Cluster Analysis 37516.1 Introduction 375Example: Public Utilities 37716.2 Measuring Distance Between Two Records 378Euclidean Distance 379Standardizing Numerical Measurements 379Other Distance Measures for Numerical Data 379Distance Measures for Categorical Data 382Distance Measures for Mixed Data 38216.3 Measuring Distance Between Two Clusters 383Minimum Distance 383Maximum Distance 383Average Distance 383Centroid Distance 38316.4 Hierarchical (Agglomerative) Clustering 385Single Linkage 385Complete Linkage 386Average Linkage 386Centroid Linkage 386Ward's Method 387Dendrograms: Displaying Clustering Process and Results 387Validating Clusters 391Two-Way Clustering 393Limitations of Hierarchical Clustering 39316.5 Nonhierarchical Clustering: The K-Means Algorithm 394Choosing the Number of Clusters (k) 396Problems 403Part VII Forecasting Time Series17 Handling Time Series 40917.1 Introduction 40917.2 Descriptive vs. Predictive Modeling 41017.3 Popular Forecasting Methods in Business 411Combining Methods 41117.4 Time Series Components 411Example: Ridership on Amtrak Trains 41217.5 Data Partitioning and Performance Evaluation 415Benchmark Performance: Naive Forecasts 417Generating Future Forecasts 417Problems 41918 Regression-Based Forecasting 42318.1 A Model with Trend 424Linear Trend 424Exponential Trend 427Polynomial Trend 42918.2 A Model with Seasonality 430Additive vs. Multiplicative Seasonality 43218.3 A Model with Trend and Seasonality 43318.4 Autocorrelation and ARIMA Models 433Computing Autocorrelation 433Improving Forecasts by Integrating Autocorrelation Information 437Fitting AR Models to Residuals 439Evaluating Predictability 441Problems 44419 Smoothing and Deep Learning Methods for Forecasting 45519.1 Introduction 45519.2 Moving Average 456Centered Moving Average for Visualization 456Trailing Moving Average for Forecasting 457Choosing Window Width (w) 46019.3 Simple Exponential Smoothing 461Choosing Smoothing Parameter alpha 462Relation Between Moving Average and Simple ExponentialSmoothing 46519.4 Advanced Exponential Smoothing 465Series With a Trend 465Series With a Trend and Seasonality 46619.5 Deep Learning for Forecasting 470Problems 472Part VIII Data Analytics20 Text Mining 48320.1 Introduction 48320.2 The Tabular Representation of Text: Document-Term Matrix and"Bag-of-Words" 48420.3 Bag-of-Words vs. Meaning Extraction at Document Level 48620.4 Preprocessing the Text 486Tokenization 487Text Reduction 488Presence/Absence vs. Frequency (Occurrences) 489Term Frequency-Inverse Document Frequency (TF-IDF) 489From Terms to Topics: Latent Semantic Analysis and TopicAnalysis 490Extracting Meaning 491From Terms to High Dimensional Word Vectors: Word2Vec 49120.5 Implementing Machine Learning Methods 49220.6 Example: Online Discussions on Autos and Electronics 492Importing the Records 493Text Preprocessing in JMP 494Using Latent Semantic Analysis and Topic Analysis 496Fitting a Predictive Model 499Prediction 49920.7 Example: Sentiment Analysis of Movie Reviews 500Data Preparation 500Latent Semantic Analysis and Fitting a Predictive Model 50020.8 Summary 502Problems 50321 Responsible Data Science 50521.1 Introduction 505Example: Predicting Recidivism 50621.2 Unintentional Harm 50621.3 Legal Considerations 508The General Data Protection Regulation (GDPR) 508Protected Groups 50821.4 Principles of Responsible Data Science 508Non-maleficence 509Fairness 509Transparency 510Accountability 511Data Privacy and Security 51121.5 A Responsible Data Science Framework 511Justification 511Assembly 512Data Preparation 513Modeling 513Auditing 51321.6 Documentation Tools 514Impact Statements 514Model Cards 515Datasheets 516Audit Reports 51621.7 Example: Applying the RDS Framework to the COMPAS Example 517Unanticipated Uses 518Ethical Concerns 518Protected Groups 518Data Issues 518Fitting the Model 519Auditing the Model 520Bias Mitigation 52621.8 Summary 526Problems 528Part IX Cases22 Cases 53322.1 Charles Book Club 533The Book Industry 533Database Marketing at Charles 534Machine Learning Techniques 535Assignment 53722.2 German Credit 541Background 541Data 541Assignment 54422.3 Tayko Software Cataloger 545Background 545The Mailing Experiment 545Data 545Assignment 54622.4 Political Persuasion 548Background 548Predictive Analytics Arrives in US Politics 548Political Targeting 548Uplift 549Data 549Assignment 55022.5 Taxi Cancellations 552Business Situation 552Assignment 55222.6 Segmenting Consumers of Bath Soap 554Business Situation 554Key Problems 554Data 555Measuring Brand Loyalty 556Assignment 55622.7 Catalog Cross-Selling 557Background 557Assignment 55722.8 Direct-Mail Fundraising 559Background 559Data 559Assignment 55922.9 Time Series Case: Forecasting Public Transportation Demand 562Background 562Problem Description 562Available Data 562Assignment Goal 562Assignment 563Tips and Suggested Steps 56322.10 Loan Approval 564Background 564Regulatory Requirements 564Getting Started 564Assignment 564References 567Data Files Used in the Book 571Index 573

O autorze

Galit Shmueli, PhD is Distinguished Professor at National Tsing Hua University's Institute of Service Science. She has designed and instructed business analytics courses since 2004 at University of Maryland, Statistics.com, The Indian School of Business, and National Tsing Hua University, Taiwan.Peter C. Bruce is Founder of the Institute for Statistics Education at Statistics.com, and Chief Learning Officer at Elder Research, Inc.Mia L. Stephens, M.S. is an Advisory Product Manager with JMP, driving the product vision and roadmaps for JMP(r) and JMP Pro(r).Muralidhara Anandamurthy, PhD is an Academic Ambassador with JMP, overseeing technical support for academic users of JMP Pro(r).Nitin R. Patel, PhD is cofounder and lead researcher at Cytel Inc. He is also a Fellow of the American Statistical Association and has served as a visiting professor at the Massachusetts Institute of Technology and Harvard University, among others.