Time Series Analysis

Time series forecasting is at the heart of some of the most important decisions in business, science, and technology — from predicting energy consumption and weather patterns to understanding financial markets and customer behavior. In this course, you’ll learn how to make sense of time-dependent data and turn it into actionable forecasts. We start with the basics: what time series data is, how it works, and why it’s different from other types of data.

You’ll explore key concepts like trends, seasonality, stationarity, and autocorrelation, and learn how to visualize and analyse time series effectively. Then we move into classical forecasting methods — including exponential smoothing and ARIMA-family models — and how to evaluate their performance using the right metrics and baselines.

Next, we introduce machine learning for time series: you’ll learn how to engineer lag features, rolling statistics, and seasonal patterns, and apply models like linear regression, decision trees, and gradient boosting. You’ll also learn how to prepare time series targets, handle non-stationarity, and combine models using ensembling strategies for improved accuracy and robustness.

Finally, we cover deep learning approaches to time series forecasting. You'll build models using recurrent architectures like RNNs, LSTMs, and GRUs, and explore convolutional and attention-based models. We’ll also look at specialized forecasting architectures such as N-BEATS, N-HiTS, and Transformers, which have set new benchmarks in recent forecasting competitions. You'll gain an understanding of when deep learning is worth the complexity, how to structure your inputs and outputs, and how to train these models on real-world time series data.

Throughout the course, we include practical examples from finance, healthcare, retail, and energy. Whether you're a data scientist, analyst, engineer, or simply curious about forecasting, this course provides a practical, end-to-end foundation for working with time series data. By the end, you'll be ready to build, evaluate, and deploy forecasting models with confidence.

Scope

The Time Series Analysis course runs for two weeks with a total workload of approximately 8-10 hours. It includes video lectures accompanied by multiple-choice self-assessments.

All learning materials and selftests are available at the course start. The first graded assignment will be released along with the course material, and the second graded assignment will be released at the end of the first week, giving learners two weeks to complete the course and submit their solutions.