Utah State University | Huntsman School of Business
This comprehensive course introduces students to state-of-the-art time series forecasting techniques, progressing from classical statistical methods to advanced deep learning architectures. Students will gain hands-on experience with real-world forecasting problems using industry-standard tools and frameworks.
Upon completion of this course, students will be able to:
- Master fundamental time series concepts and decomposition techniques
- Implement classical forecasting methods (ETS, ARIMA/SARIMA)
- Apply machine learning algorithms to time series problems
- Design and train deep neural networks for sequence modeling
- Deploy production-ready forecasting models at scale
- Evaluate and compare model performance using appropriate metrics
- Programming: Basic Python proficiency (variables, loops, functions)
- Mathematics: College-level statistics and linear algebra
- Software: Google account for Colab access (no local installation required)
For students needing a refresher, we provide a comprehensive Python Crash Course covering:
- Python basics, NumPy, Pandas
- Data visualization (Matplotlib, Seaborn)
- Time series data manipulation
- Time series components and patterns
- Stationarity and transformations
- Autocorrelation and partial autocorrelation
- Python environment configuration
- Essential libraries and tools
- Google Colab setup and best practices
- Simple, Holt's, and Holt-Winters methods
- ETS (Error, Trend, Seasonal) models
- Model selection and validation
- AR, MA, and ARMA processes
- ARIMA and seasonal ARIMA (SARIMA)
- Box-Jenkins methodology
- Feature engineering for time series
- Tree-based methods (Random Forest, XGBoost, LightGBM)
- Cross-validation strategies
- Feedforward networks for time series
- Backpropagation and optimization
- TensorFlow/Keras implementation
- Recurrent Neural Networks (RNN)
- Long Short-Term Memory (LSTM) networks
- Bidirectional and stacked architectures
- Forecasting at scale
- Handling seasonality and holidays
- Uncertainty quantification
- PyCaret: AutoML for time series
- TensorFlow/Keras: Deep learning
- Prophet/NeuralProphet: Scalable forecasting
- Streamlit: Interactive dashboards
- Google Colab: Cloud-based Jupyter notebooks
- GitHub: Version control and collaboration
- Requirements: Modern web browser, stable internet connection
The course includes various real-world datasets:
- Airline Passengers: Classic time series dataset
- Retail Sales: Rossmann store sales data
- Economic Indicators: US GDP and macroeconomic data
- Stock Market: Financial time series examples
- Custom Projects: Students can bring their own data
All datasets are available in the data/ directory.
- Click on any notebook's "Open in Colab" button
- Sign in with your Google account
- Run cells sequentially (Shift+Enter)
# Clone the repository
git clone https://github.com/PJalgotrader/Deep_forecasting-USU.git
cd Deep_forecasting-USU
# Create virtual environment (recommended)
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Launch Jupyter
jupyter notebookNote: The installation may take several minutes as it includes deep learning frameworks and multiple ML libraries. For lighter installation, you can install packages as needed for specific modules.
- Forecasting: Principles and Practice by Hyndman & Athanasopoulos
- Deep Learning by Goodfellow, Bengio, and Courville
- Course papers in
Lectures and codes/
Pedram Jahangiry, PhD, CFA
Professional Practice Assistant Professor
Data Analytics and Information Systems
Huntsman School of Business, Utah State University
- 📧 Email: [email protected]
- 📺 YouTube Channel
- 🐦 Twitter/X
Office Hours: By appointment
Dr. Jahangiry brings extensive industry experience from his role as a Research Associate in the Financial Modeling Group at BlackRock NYC. His research focuses on machine learning, deep learning, and time series forecasting applications in finance and business analytics. He mentors students at the Analytics Solutions Center, providing hands-on experience with real corporate analytics projects.
We welcome contributions from students and the community! Please feel free to:
- Report issues or bugs
- Suggest improvements or new examples
- Share your projects and applications
- Submit pull requests with enhancements
This course material is freely available for educational purposes. All rights reserved by Dr. Pedram Jahangiry and Utah State University.
Special thanks to:
- The Huntsman School of Business for supporting this course
- The Analytics Solutions Center team
- All students and contributors who have helped improve this material
- The open-source community for the amazing tools and libraries
Fall 2025 | Utah State University
Empowering the next generation of data scientists and forecasting experts