Tepper · MBA

45-851 Data Mining

Finding structure in messy business data, then deciding whether to trust it.

Tepper's MBA data mining course, which I have taught six times since 2023. It covers clustering, PCA, regression, classification, forecasting, and text, with labs in Python, and it keeps asking what each model is for. Labs are in Python. The goal is not a long list of algorithms. It is a workflow students can apply to an unfamiliar dataset the week after the course ends.

Topics

1. Data exploration and visualization in Python
2. Clustering, with a lab
3. Principal component analysis
4. Regression, with a lab
5. Classification, with a lab
6. Forecasting with time series data
7. Natural language processing, with a lab
8. Neural networks and large language models

How the course works

The course is organized around business questions, not algorithms. Each module opens with one a manager would ask: which customers look alike, what a house will sell for, who is about to leave, what customers are saying in their reviews. The first class is a case from my own work, predicting drug price spikes for a 40-hospital health system that buys $1.5 billion of pharmaceuticals a year, followed from the business question to a commercial product.

Students work the way analysts work now. Labs are in Python in Colab, generative AI is allowed, and teams hand in their AI chat transcripts with the work. The classification project is a prediction competition: every team scores a holdout file, I grade the predictions against the true labels, and the top fifth earn extra credit.

Final projects

In the final project, teams bring their own business question and data, often from their own employer, and answer it with a method from the course. Projects from Spring 2024, Fall 2024, and Spring 2025, grouped by type, with one example each.

• Finance, markets, and credit risk6
  Do a CEO's posts move an electric carmaker's stock price?
• Sports analytics5
  Predicting baseball Hall of Fame induction from career statistics
• Healthcare and wellbeing5
  Does wearing a fitness tracker go with better health?
• Customer segmentation and retail4
  Segmenting 85 grocery stores by their department sales mix
• Transportation and mobility4
  Where should Pennsylvania put new EV chargers?
• Churn, retention, and satisfaction3
  Predicting and segmenting airline passenger satisfaction
• Real estate and places3
  What drives Airbnb nightly prices in New York?
• Economic forecasting2
  Forecasting US retail sales with Prophet
• Media and entertainment2
  What predicts a song's stream count?

How the course developed

I first taught Data Mining in Fall 2023, as an adjunct, to full-time MBA students. I taught it every fall and spring through Fall 2025, in both the full-time and the online hybrid programs.

The first version was in R. For the Spring 2024 online hybrid section I scripted and recorded a video module for each topic. In Fall 2024 I moved the whole course to Python and then re-recorded the videos, and Fall 2025 added principal component analysis.

Offerings

• Fall 2023 full-time
• Spring 2024 online hybrid
• Fall 2024 full-time
• Spring 2025 online hybrid
• Fall 2025 full-time
• Fall 2025 online hybrid

Enrolled students get the notebooks and video walkthroughs on Canvas.