Data Driven Modeling - Day 2
Repository: https://github.com/colbrydi/data_driven_modeling
Agenda (80 minutes)
- 10 min - Check-in and review of Part 1: what did statistics tell us?
- 20 min - Introduce supervised and unsupervised learning, train/test splits, and overfitting
- 35 min - Guided work in
Assignment_Instructions_Part_2.ipynb - 15 min - Share evidence about dataset size and the number of features
Learning Goals
By the end of this class, you should be able to:
- Distinguish regression, classification, and clustering using the same dataset.
- Use evidence from your own experiments to judge whether a dataset is large enough for a model.
In-Class Focus
Statistics describe; machine learning predicts. Today we use scikit-learn on one small dataset to ask three different questions, compare each model against a simple baseline, and watch what happens when a model sees too few observations or too many features.
Use the repository as the primary working reference:
- Repository: https://github.com/colbrydi/data_driven_modeling
- Notebook:
Assignment_Instructions_Part_2.ipynb
The notebook gives you a working scaffold and expects you to look things up in the scikit-learn documentation. As you work, consider whether the model beats the baseline, how you know you are not fooling yourself, and whether machine learning was necessary at all.
Before Next Class
- Finish any remaining work in
Assignment_Instructions_Part_2.ipynb, including the reflection. - Write a short description of one dataset from your own work or a public source: number of observations, number of features, and the prediction question you would ask. Bring it to class.
- Preview
Assignment_Instructions_Part_3.ipynb.