Hands-On Prescriptive Analytics: Optimizing Your Decision Making with Python

October 2024: First EditionRevision History for the First Edition2024-10-16: First ReleaseBusiness decisions in any context—operational, tactical, or strategic—can have considerable consequences. Whether the outcome is positive and rewarding or negative and damaging to the business, its employees, and stakeholders is unknown when action is approved. These decisions are usually made under the proverbial cloud of uncertainty.With this practicalguide, data analysts, data scientists, and business analysts will learn why and how maximizing positive consequences and minimizing negative ones requires three forms of rich information: Descriptive analytics explores the results from an action—what has already happened. Predictive analytics focuses on what could happen. The third, prescriptive analytics, informs us what should happen in the future.While all three are important for decision-makers, the primary focus of this book is on the third: prescriptive analytics. Author Walter R. Paczkowski, Ph.D. shows you:The... Preface The Book’s Topic The Book’s Audience What You Will Learn–and How to Apply It The Book’s Structure Conventions Used in This Book O’Reilly Online Learning How to Contact Us Acknowledgments I. Introduction and Background 1. An Analytical Framework A Decision-Making Framework The Analytics Evolution The Data Science Split The Information Glue How the Analytics Fit Together Descriptive Analytics Predictive Analytics Forms of Predictive Analytics Menu of prediction options Creation of the menus Natural Projects What-if projects Examples of decision menus Vignette 1: Operational scale-view Vignette 2: Tactical scale-view Vignette 3: Strategic scale-view Uncertainty: Multiple Sources and Problems Uncertainty Versus Risk The Menu Cost of Uncertainty Probabilities and Uncertainty Prescriptive Analytics as a Separate Discipline Decision-Makers’ Options Business Goals and Constraints Key Decision Question: What Is the Best Decision? The Analytics Flow Prescriptive Analytics and Decision Making Summary 2. Prescriptive Methods: Overview Introduction to Prescriptive Analytics Methods Proprietary Methods Public Domain Methods Summary of Prescriptive Analytics Methods Methods Reviewed in the Literature Probabilistic models Machine learning and data mining Mathematical programming Evolutionary computation Simulations Logic-based models Three Categories of Methods Mathematical programming Simulation Decision trees Umbrella Classes: Non-Stochastic and Stochastic Definition of Non-Stochastic and Stochastic Methods Examples Based on Scale-Views The Role of Operations Research Summary II. Essential Background Material 3. Python Essentials Python Structure: Overview Worldwide Community Extensive Array of Packages Easy to Use Operating Systems A Programming Language It Is Free Python Basics Naming Conventions Python Data Structures Tuples Lists Dictionaries Iterables Basic Python Operators Introduction to Built-In Functions enumerate round print range Introduction to User-Defined Functions Positional arguments Named arguments Conditional Statements: if-else Python Looping Constructs for loops while loops Python Packages Data Management NumPy pandas Data Visualization Statistical Analysis and Modeling statsmodels scikit-learn SciPy Working with Python Packages Using Anaconda Updating Python Packages Installing Packages Importing Packages Go-To References Summary 4. Probability Essentials The World Is Ruled by Probabilities What Are Probabilities? Fundamental Probability Concepts Frequency-Based Probabilities Counting Functions Independence and Conditional Probability Summary of Probability Rules Limit Definition of Probabilities Subjective-Based Probabilities: Introduction Bayes’ Theorem: Derivation Bayes’ Theorem: Python Implementation Probability Distributions: Overview Three Basic Probability Distributions: Binomial, Uniform, and Normal Binomial distribution Uniform distribution Normal distribution Key Distribution Parameters Summary III. Non-Stochastic Prescriptive Analytic Methods 5. Mathematical Programming: Overview Background Reason for Popularity The Objective Function: Prescriptive Analytics Goal Linear Programming Technical Overview The simplex approach Digression: The production function Other solution methods Handling special issues Menus and Linear Programming Python Use-Cases Operational scale-view: Example 1 Operational scale-view: Example 2 Integer Programming Technical Overview Menus and Integer Programming Python Use-Cases Tactical scale-view Case 1: Select as many as needed Case 2: Select a specific number Case 2a: Select only one Case 3: Select based on logical conditions Strategic scale-view Case 1: Select as many as needed Case 2: Select a specific number Case 2a: Select only one Case 3: Select based on logical conditions Mixed Integer Programming Technical Overview Python Use-Case Summary 6. Decision Tree Analysis: Overview Extending the Menu into Time Introduction to Decision Trees Clarification of Decision Trees Background Use of Trees in Decisions Comparing the Two Decision Trees Python Use-Case Use-Case Background Use-Case Detailed Data Role of the DADT DADT Analysis Reaching a Decision Using a DADT Summary DADT Function IV. Stochastic Prescriptive Analytic Methods 7. Simulation Essentials What Is a Simulation? The Simulation Age Types of Simulations Non-stochastic Stochastic Non-Stochastic Simulations: The Process Stochastic Simulations: The Process Pseudo-Random Number Generators: A Brief Introduction Linear congruential generator Mersenne Twister generator Generating random numbers in Python random NumPy Simulation Models: Overview KPMs to Measure Aggregation Methods The Need for Stochastic Simulations The Extent of Uncertainty Randomness and the Degree of Uncertainty Summary 8. Simulation Examples Example 1: Coin Toss Example 2: Die Toss Example 3: Regression Analysis Example 4: Mathematical Programming Example 5: Decision Tree Summary 9. Developing Menu Options The Nature of What-If Questions Menu Generating Questions: A Deep Dive The Structure of What-If Questions What-If Analysis Versus Sensitivity Analysis Non-Stochastic Use-Cases Pricing What-If Analysis: Basic The pricing problem Predictive model estimation What-if analyses Pricing What-If Analysis: Advanced The pricing problem Predictive model estimation What-if analyses Stochastic Use-Case: Synthetic Data Specifying the Process to Simulate Example of a System’s Process Flow The what-if questions A diffusion model Generating synthetic data What-if analysis with the synthetic data Summary 10. Developing Menu Priors Background Digression on Beliefs and Priors Developing Probability Weights Eliciting Probability Distributions of Beliefs Elicitation Method 1: Experimental Design-Based A conjoint analysis Discrete choice analysis MaxDiff analysis Methods’ problem Elicitation Method 2: Direct Questioning Elicitation Method 3: Activities Analyzing Elicited Probability Distributions Elicitation Analysis 1: Averaging Elicitation Analysis 2: CoDA Elicitation Analysis 3: Bootstrapping Python Use-Case Elicitation Example Analysis 1: Averaging Elicitation Example Analysis 2: CoDA Elicitation Example Analysis 3: Bootstrapping Summary 11. One-Time Decisions Evidence of the Problem Sequential Decisions: Introduction Sequential Decisions 1: The Business Case Sequential Decisions 2: Post-Business Case Sequential Analysis: Advanced Framework Markov Decision Problem Simulations and Reinforcement Learning Automating Sequential Decision Making Summary Glossary Bibliography Index Business decisions in any context—operational, tactical, or strategic—can have considerable consequences. Whether the outcome is positive and rewarding or negative and damaging to the business, its employees, and stakeholders is unknown when action is approved. These decisions are usually made under the proverbial cloud of uncertainty. With this practical guide, data analysts, data scientists, and business analysts will learn why and how maximizing positive consequences and minimizing negative ones requires three forms of rich information: Descriptive analytics explores the results from an action—what has already happened. Predictive analytics focuses on what could happen. The third, prescriptive analytics , informs us what should happen in the future. While all three are important for decision-makers, the primary focus of this book is on the third: prescriptive analytics. Author Walter R. Paczkowski, Ph.D. shows you: The...