book

Analytical Skills for AI and Data Science

by Daniel Vaughan

May 2020

Beginner to intermediate

242 pages

7h 17m

English

O'Reilly Media, Inc.

Read now

Unlock full access

Why Analytical Skills for AI?Use Case-Driven ApproachWhat This Book Isn’tWho This Book Is ForWhat’s NeededConventions Used in This BookUsing Code ExamplesO’Reilly Online LearningHow to Contact UsAcknowledgments
What Is AI?Why Current AI Won’t Deliver on Its PromisesHow Did We Get Here?The Data RevolutionA Tale of Unrealized ExpectationsAnalytical Skills for the Modern AI-Driven EnterpriseKey TakewaysFurther Reading
Descriptive, Predictive, and Prescriptive QuestionsWhen Predictive Analysis Is Powerful: The Case of Cancer DetectionDescriptive Analysis: The Case of Customer ChurnBusiness Questions and KPIsKPIs to Measure the Success of a Loyalty ProgramAn Anatomy of a Decision: A Simple DecompositionAn Example: Why Did You Buy This Book?A Primer on CausationDefining Correlation and CausationSome Difficulties in Estimating Causal EffectsUncertaintyUncertainty from SimplificationUncertainty from HeterogeneityUncertainty from Social InteractionsUncertainty from IgnoranceKey TakeawaysFurther Reading
From Business Objectives to Business QuestionsDescriptive, Predictive, and Prescriptive QuestionsAlways Start with the Business Question and Work BackwardFurther Deconstructing the Business QuestionsExample with a Two-Sided PlatformLearning to Ask Business Questions: Examples from Common Use CasesLowering ChurnCross-Selling: Next-Best OfferCAPEX OptimizationStore LocationsWho Should I Hire?Delinquency RatesStock or Inventory OptimizationStore StaffingKey TakeawaysFurther Reading
Understanding What Is ActionablePhysical LeversHuman LeversWhy Do We Behave the Way We Do?Levers from RestrictionsLevers That Affect Our PreferencesLevers That Change Your ExpectationsRevisiting Our Use CasesCustomer ChurnCross-SellingCapital Expenditure (CAPEX) OptimizationStore LocationsWho Should I Hire?Delinquency RatesStock OptimizationStore StaffingKey TakeawaysFurther Reading
Why Do We Need to Simplify?First- and Second-Order EffectsExercising Our Analytical Muscle: Welcome FermiHow Many Tennis Balls Fit the Floor of This Rectangular Room?How Much Would You Charge to Clean Every Window in Mexico City?Fermi Problems to Make Preliminary Business CasesRevisiting the Examples from Chapter 3Customer ChurnCross-SellingCAPEX OptimizationStore LocationsDelinquency RatesStock OptimizationStore StaffingKey TakeawaysFurther Reading
Where Does Uncertainty Come From?Quantifying UncertaintyExpected ValuesMaking Decisions Without UncertaintyMaking Simple Decisions Under UncertaintyDecisions Under UncertaintyIs This the Best We Can Do?But This Is a Frequentist ArgumentNormative and Descriptive Theories of Decision-MakingSome Paradoxes in Decision-Making Under UncertaintyThe St. Petersburg ParadoxRisk AversionPutting it All into PracticeEstimating the ProbabilitiesEstimating Expected ValuesFrequentist and Bayesian MethodsRevisiting Our Use CasesCustomer ChurnCross-SellingCAPEX OptimizationStore LocationsWho to HireDelinquency RatesStock OptimizationKey TakeawaysFurther Reading
What Is Optimization?Numerical Optimization Is HardOptimization Is Not New in Business SettingsPrice and Revenue OptimizationOptimization Without UncertaintyCustomer ChurnCross-SellingCAPEX InvestmentOptimal StaffingOptimal Store LocationsOptimization with UncertaintyCustomer ChurnCross-SellingOptimal StaffingTricks for Solving Optimization Problems Under UncertaintyKey TakeawaysFurther Reading
Analytical SkillsAsking Prescriptive QuestionsUnderstanding CausalitySimplifyEmbracing UncertaintyTackling OptimizationThe AI-Driven Enterprise of the FutureBack to AISome Final Thoughts
What Is Machine Learning?A Taxonomy of ML ModelsSupervised LearningUnsupervised LearningSemisupervised LearningRegression and ClassificationMaking PredictionsCaveats to the Plug-in ApproachWhere Do These Functions Come From?Making Good PredictionsFrom Linear Regression to Deep LearningLinear RegressionNeural NetworksA Primer on A/B TestingFurther Reading

Content preview from Analytical Skills for AI and Data Science

Chapter 3. Learning to Ask Good Business Questions

Chapter 2 provided a quick overview of the general framework we’ll be developing in the upcoming chapters. Since our ultimate objective is to translate business problems into prescriptive solutions, we should start learning how to ask the right questions. I hope it doesn’t come as a surprise that learning to frame the questions can have an impact comparable in magnitude to adopting the techniques that will follow.

We also introduced a very simple technique that I’ve found quite useful to understand what we really want to accomplish: the sequence of why questions. You start by questioning what you think you are trying to accomplish, then move up one level or stop when you are convinced that the business objective is in fact just right. In our voyage to find prescriptive solutions, it is of the utmost importance to guarantee that we are tackling the right objectives. One nice byproduct that will be quite handy in Chapter 4 is that this usually enlarges the set of possible actions or levers we have. This is usually the case when we start by questioning an action and the procedure ends up taking us to the metrics we really want to affect. It is almost natural, then, to question if there are other actions that can be used to impact the same objective.

In this chapter, we will delve a bit more into some of the best practices when asking good business questions; understand the difference between descriptive, predictive, and prescriptive ...