In Week 8, we fit a model to a cross-section of customers. This week, the model considers time and how we can predict the future. We cover some of the most common forecasting approaches used in the wild, rank them with a backtest, and hedge the number you report.
There is no new Lab or Homework this week, so the runway is clear to finish Project 2. Check Canvas for the due date.
Self-paced deck. The SWAG baseline every method has to beat, the backtest that ranks the methods, the forecasting approaches from judgment through statistics, the leaderboard, and how to size the hedge you put around the number. Every slide carries an in-deck AI tutor and a self-check.
Three trainers you run in the browser. In the Forecast-Off you commit your own SWAG first, then watch eight methods get scored on a backtest. In the marketing-driver trainer you fit sales on spend and see where the line stops being trustworthy. In the cohort-retention trainer you watch attrition erode a cohort's revenue over time. Each one checks your work and carries the AI tutor.
The week is the work window for Project 2, the credit-default classifier you built on the Week 8 model. Hold out data, measure on the rows the model never saw, and state a result you can defend. Same discipline this week preached, a different model.
The class AI is built into every page here. Open the Ask AI panel on the lecture, the launchpad, or any trainer. It can read what you have typed and nudge you toward the fix without handing you the answer.