Interpretable Machine Learning for Finance

Northwestern University

MLDS 490-0-1 Special Topics

Syllabus and Data

Syllabus and data sets will be available on the course Canvas.

Textbooks

Students are highly encouraged to purchase the following two books for the course:

• Molnar, Christoph. Interpretable Machine Learning: A Guide for Making Black Box Models Explainable. 3rd ed. https://christophm.github.io/interpretable-ml-book/. Accessed on March 28, 2025.

• Molnar, Christoph. Interpreting Machine Learning Models With SHAP: A Guide With Python Examples and Theory on Shapley Values. Independent publication, 2023.

Key Repo Files

• conda_env_iml4finance_2026.yml : cross-platform conda environment for the course

• IML4Finance.code-workspace: workspace file for the local repository

• .vscode/Interpretable ML.code-profile: VS code profile of extensions and keyboard shortcuts

System Prerequisites

• Miniforge3

• VS Code IDE

• Git

• GitHub Desktop

Clone the Repo

1. Open GitHub Desktop (and link your GitHub account).
2. Click on File > Clone Repository....
3. Select the URL tab.
4. Enter the URL of the repository
5. Click on Choose... and select the directory where you want to save the repository.
6. Click on Clone.

Conda Environment

1. Open a terminal.

   On Windows, use Miniforge Prompt or a shell where conda is initialized.

   On macOS or Linux, use your regular terminal.

2. Change the directory to the local repository (see Step 5 in the previous section).

3. Create the conda environment from conda_env_iml4finance_2026.yml:

conda env create -f conda_env_iml4finance_2026.yml

conda_env_iml4finance_2026.yml is the only supported environment file for this repo.

4. Activate the conda environment:

conda activate env_iml4finance_2026

5. Launch VS Code:

code

Lecture, lab, and rendered output filenames use underscore-based names such as Lecture_02.qmd, Lab_01.qmd, and Lecture_02.html to avoid shell and Quarto issues caused by spaces in file paths.

Repository Structure

• Lectures/: lecture source files (.qmd), rendered lecture outputs (.html, _files/), lecture PDFs, lecture model folders, Images/, and references.bib

• Labs/: lab source files (Lab_01.qmd through Lab_04.qmd), rendered lab outputs (.html, _files/), EDA reports, and lab model folders

• Quizzes/: quiz scripts and quiz-specific data files

• Examples/: standalone example spreadsheets used in the course

• course_utils/: shared Python utilities for labs and other course materials

• Data/: dataset-specific loaders, documentation, and local data assets

Labs now import shared utilities from course_utils.helpers, while lectures continue to use local lecture assets from Lectures/.

Rendered Quarto outputs now live beside their source .qmd files in Labs/ and Lectures/. Model folders created by the course materials also live beside the .qmd files that generate or consume them.

VS Code Profile

1. Open VS Code
2. Type Ctrl + Shift + P on Windows and Cmd + Shift + P on Mac
3. Type >Profiles: New Profile
4. Delete the profile that is called "Untitled"
5. Click on the drop-down arrow next to the "New Profile" button
6. Click Import Profile
7. Click Select File
8. Choose Interpretable ML.code-profile

Hidden folders on a Mac

To see hidden folders (that start with a dot) on a Mac, toggle on/off with Cmd + Shift + ..