Machine Learning

Lecture slides, Python notebooks, data and cheatsheets for the machine learning courses taught by Prof. Pedram Jahangiry at Utah State University.

Every notebook runs in three ways: on Google Colab with nothing to install, or on your own computer with uv (recommended) or conda. See Setup.

📂 What is in this repository

Folder	Content
Lectures and codes	One folder per module: the lecture slides (PDF) and the Python notebooks
Platforms and tools	Google Colab jumpstart, the PyCaret demos (regression, classification, time series), and the uv quick start (cheat sheet + ten-second test)
data	The datasets used in the notebooks
Cheatsheets	Python, pandas, statistics, probability and machine learning cheatsheets

🎲 Topics covered

Module	Topic	Notebooks use
1	Introduction to Machine Learning	slides only
2	Setting up the Machine Learning Environment (Python crash course, EDA)	pandas, NumPy, Matplotlib, seaborn
3	Linear Regression (Econometrics approach)	statsmodels
4	Machine Learning Fundamentals	slides only
5	Linear Regression (Machine Learning approach)	scikit-learn, PyCaret
6	Penalized Regression (Ridge, LASSO, Elastic Net)	scikit-learn, PyCaret
7	Logistic Regression	scikit-learn, PyCaret
8	K-Nearest Neighbors (KNN)	scikit-learn, PyCaret
9	Classification and Regression Trees (CART)	scikit-learn, PyCaret
10	Bagging and Boosting (Random Forest, AdaBoost, XGBoost, LightGBM, CatBoost)	scikit-learn, xgboost, PyCaret
11	Dimensionality Reduction (PCA)	scikit-learn, pca
12	Clustering (KMeans, Hierarchical, K-modes)	scikit-learn, kmodes, PyCaret
Extra	Support Vector Machines (in miscellaneous)	slides only

Setup

Read this first: PyCaret is now pycaret-core

The second half of each module uses PyCaret, a low-code machine learning library. The original pycaret package is no longer maintained and does not run on current Python versions, including the one on Google Colab. If you pip install pycaret you get one of these:

• RuntimeError: Pycaret only supports python 3.9, 3.10, 3.11 ... Please DOWNGRADE
• a long failed build of NumPy

This course uses pycaret-core, the community-maintained continuation of PyCaret 3 under the sktime organization. You still write import pycaret, and every function used in the lectures (setup, compare_models, create_model, tune_model, plot_model, ...) works the same. Only the install command changed:

pip install pycaret-core lightgbm xgboost catboost shap

lightgbm, xgboost and catboost are listed so that compare_models() shows the same models for everybody. shap is needed by interpret_model.

Option 1: Google Colab (nothing to install)

1. Open a notebook on GitHub and click the Open in Colab badge at the top.
2. Save a copy to your own Google Drive if you want to keep your changes.
3. Run the cells from top to bottom. The PyCaret notebooks contain an install cell like this one. It installs the PyCaret stack on Colab and does nothing anywhere else:

import sys
if "google.colab" in sys.modules:
    !pip install -q pycaret-core lightgbm xgboost catboost shap

Start from a fresh runtime (Runtime > Disconnect and delete runtime) if you already installed or imported anything PyCaret-related in that session. No runtime restart is needed after the install.

Option 2: your own computer with uv (recommended)

uv installs Python and all the packages for you, in a folder inside this repository. It does not touch any Python or Anaconda you already have. New to uv? Start with Platforms and tools/uv/: a quick start, the conda-to-uv cheat sheet, and a ten-second test project. The same folder, and the same commands, are used in the Deep Learning and Deep Forecasting courses.

1. Install uv (instructions), then reopen the terminal and check uv --version.

   Windows (PowerShell):

   powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
   

   macOS / Linux:

   curl -LsSf https://astral.sh/uv/install.sh | sh
   

2. Get the repository (clone into a normal local folder, not inside Google Drive or OneDrive) and build the environment:

   git clone https://github.com/PJalgotrader/Machine_Learning-USU.git
   cd Machine_Learning-USU
   uv sync
   

   Optional first: the ten-second test in Platforms and tools/uv/simple_test (uv sync, then uv run python main.py) prints your Python and pandas versions and uv is working!.

3. Check it:

   uv run python scripts/check_environment.py
   

   The last line should say Your course environment is ready.

4. Start Jupyter:

   uv run jupyter lab
   

VS Code: register the environment as a Jupyter kernel once, then pick it with Select Kernel > Jupyter Kernel:

uv run python -m ipykernel install --user --name ml-course --display-name "Python 3.13 (ML course)"

Option 3: your own computer with conda

git clone https://github.com/PJalgotrader/Machine_Learning-USU.git
cd Machine_Learning-USU
conda env create -f environment.yml
conda activate ml_pycaret
python scripts/check_environment.py
jupyter lab

In VS Code, the ml_pycaret environment shows up under Select Kernel > Python Environments.

What the environment contains

Python 3.13 with pycaret-core, scikit-learn, pandas, NumPy, statsmodels, Matplotlib, seaborn, LightGBM, XGBoost, CatBoost, shap, kmodes, pca, yfinance and JupyterLab. The exact versions are locked in uv.lock. The package list lives in pyproject.toml (uv) and environment.yml (conda).

The automated EDA notebook in Module 2 (pandas-profiling, sweetviz, dtale) is not part of this environment. Run that one on Colab.

Troubleshooting

Symptom	Fix
Pycaret only supports python 3.9, 3.10, 3.11 ... Please DOWNGRADE	You installed the old pycaret package. Colab: delete the runtime and run the notebook's install cell. Local: pip uninstall pycaret, then build the environment as above.
pip tries to build NumPy from source and fails	Same cause: the old pycaret package on a new Python. Install pycaret-core instead.
ValueError: Estimator xgboost Not Available (or catboost, lightgbm) on Colab	The library was installed after PyCaret was already imported. Runtime > Disconnect and delete runtime, then run the notebook from the top.
The checker reports the wrong Python version	The notebook or terminal is not using the course environment. Use uv run ..., or conda activate ml_pycaret, or pick the right kernel in VS Code.
DLL load failed on Windows inside a uv environment	The environment was built on top of Anaconda's Python. Delete the .venv folder and run uv sync again. This repository tells uv to use its own Python.
Install errors about paths that are too long (Windows)	Clone the repository into a short path such as C:\ML, and not inside OneDrive or Google Drive.
ETSResults.simulate() got an unexpected keyword argument 'random_state' in the time series demos	statsmodels is too new for the time series module. The time series notebooks and the local environment pin statsmodels<0.15.

🚀 About Me

Pedram Jahangiry, CFA, is a Professional Practice Associate Professor of Data Analytics and Information Systems in the Huntsman School of Business at Utah State University. Prior to joining the Huntsman School, Pedram was a research associate within the Financial Modeling Group at BlackRock NYC. His current research is involved in machine learning, deep learning and time series forecasting. Pedram is one of the project mentors at the Analytics Solutions Center, where students from across USU’s Logan and Statewide campuses work with corporate partners on analytics projects.

🔗 Links