Course: MGT 6480 — Predictive Analytics, Mod IV, Spring 2026

Instructor: Sina Khorasani (PhD), Assistant Professor of Operations Management, Owen Graduate School

Optional textbook anchor: An Introduction to Statistical Learning with Applications in Python (James, Witten, Hastie, Tibshirani, Taylor, 2023)

Worked examples: Gabriel's three numbered assignments + the Customer Retention group case + the B2B Churn Agent capstone

This masterclass is the long-form synthesis of the course — built so a careful reader who has never opened a Jupyter notebook can finish it understanding why predictive analytics matters strategically, how each model in the canon actually works, how to read the Python output, and how to translate model performance into a business case an executive will sign off on. It is heavy on intuition, heavy on worked examples, and built to be re-readable years from now.

How to read this masterclass

Read in order. Each module assumes you understood the one before. The narrative arc is deliberate:

• Modules 1–2 set the strategic stakes. Why prediction matters, when to use ML, how to spot an AI project that will fail before you waste money on it.
• Module 3 is your forensic kit — every model that follows depends on understanding the data first.
• Modules 4–9 walk the model canon: linear regression → logistic regression → cross-validation → shrinkage → SVMs → trees and ensembles. Every model is taught the same way: the geometric intuition, the math you actually need, the Python output, the failure modes, and the worked example from Gabriel's coursework.
• Module 10 is the most strategically important section. Causation, recommendation systems, selection bias, and the "from prediction to decision" leap that separates analytics from business impact.
• Module 11 zooms back out: how AI is reshaping consulting, where the agentic era is heading, and how Gabriel's group capstone (a working B2B Churn Agent) brings everything together.
• The Master Playbook is the cross-class diagnostic kit — when to use which model, what to look for in the residuals, how to present to an executive, what red flags to escalate.

How this masterclass connects to the rest of the library

Predictive Analytics sits at the intersection of several other masterclasses:

• Managerial Statistics — the foundational sibling of this masterclass. Where Predictive Analytics starts (linear and logistic regression, cross-validation, ML model canon), Managerial Statistics is what came before: descriptive statistics, sampling theory, the CLT, confidence intervals, hypothesis testing, A/B testing, simple and multiple regression with diagnostics, and the data-ethics framework. The Managerial Statistics OSHA case (LASSO vs. kitchen-sink with overfitting diagnostic) is the bridge into Module 7 here on shrinkage. The Managerial Statistics QWE customer-churn case is the bridge into Module 5 here on classification. The Managerial Statistics Home Credit ethics case is the bridge into Module 10 here on causation and selection bias. Read Managerial Statistics first if you need the inferential foundations; read Predictive Analytics next for the technical depth and the modern Python-based ML canon.
• Managerial Decision Making — predictive analytics is the technical machinery that produces the inputs for decisions. The decision frameworks live there; the prediction tools live here.
• Strategic Management of Technology — when AI becomes a competitive moat (BenchSci, Google search, VideaHealth), and how to think about it strategically.
• Operations Management — demand forecasting, inventory optimization, the operational use cases for prediction.
• Service Operations — customer churn, retention, service-quality prediction, the operational substrate for the Customer Retention group case.