Time series data is essential in fields like finance, energy, healthcare, and climate science. This introductory course provides a solid foundation in the tools and techniques for analyzing, forecasting, and detecting anomalies in time series data. Participants will gain skills to uncover patterns, make predictions, and identify unusual behavior: capabilities that are crucial across many applications.
The course begins with exploratory data analysis and fundamental concepts such as stationarity and autocorrelation. From there, participants will learn to build and interpret ARIMA models, one of the most widely used classical forecasting methods. We also introduce multivariate time series forecasting and explore how machine learning methods like XGBoost can be adapted for sequential data through feature engineering.
Additionally, the course covers the basics of anomaly detection and change point detection, helping participants recognize system failures, outliers, or shifts in data behavior.
Through hands-on exercises and real-world examples, participants will develop practical skills for time series analysis. This course is ideal for those new to time series or looking to solidify their foundational knowledge before advancing to more complex methods.
An advanced course covering deep learning approaches and advanced forecasting/anomaly detection techniques is offered separately.
Learning objectives
- Understand Time Series Fundamentals.
- Preprocess Data and Perform Feature Engineering.
- Train and Evaluate Forecasting Models (Classical and ML).
- Understand Anomaly Detection and Change Point Detection.
- Apply Methods to Real Data and Avoid Common Pitfalls.
