Week 12: Other Techniques — imbalanced and time-series data
Learning Objectives
By the end of this week, students will be able to:
- Identify class imbalance and explain why it distorts standard metrics
- Apply resampling strategies (oversampling, undersampling, SMOTE) and class weighting
- Describe the structure of time-series data and why standard train/test splits fail
- Apply a time-respecting validation strategy for temporal data
Perspectival Reading
Reading: TBD
Reflection Questions
- Imbalanced datasets often reflect a world where certain events are rare but high-stakes. What is lost when we “balance” them artificially?
- Who typically occupies the minority class in socially consequential ML problems (fraud detection, medical diagnosis)?
- Time-series models are trained on the past to predict the future. What assumptions does that embed about how the world changes?
Slides
Notebook Demo
Open in Google Colab (link TBD)
Lab Assignment
Week 12 Lab — GitHub Classroom (link TBD)