Goal: How does time series analysis differ from other machine learning models?

• Read the Chapter 2. Time series graphics, Chapter 7. Time series regression models and Chapter 9. ARIMA models from Forecasting: principles and practice to get the general idea of core concepts related to time series analysis.
  • The textbook contains code in R, so you don't need to pay much attention to it - we'll be doing analogous things in Python during the lab.

1. Instead of a lecture, we go straight to the Q&A session with a short series of questions to “warm up” (based on the textbook). Also, in the hands-on session, the most important content is summarized alongside the practical activities.
2. Q&A Session:
   1. How time series data differ from other data?
   2. Explain the terms: trend, seasonal, cyclic
   3. What is the difference between stationary and non-stationary time series data?
   4. What are the AR (Autoregressive) and MA (Moving Average) models?
   5. When we develop a machine learning model, we need some input feature matrix (X) and some vector (y) we want to predict. This is a bit tricky when it comes to time series. Which features can we include for this purpose?
3. Practice session:
   1. Today's lab is placed in one Jupyter Notebook: Time series analysis
4. Advanced practice session:
   1. If you want to evaluate more advanced models, go to the optional Advanced section in the notebook.
5. Report:
   1. Send the final version of the notebook (to download it, click in Google Colab: File > Download > Download .ipynb). Make sure you have done all required 6 tasks.

• R.J. Hyndman & G. Athanasopoulos, G. – Forecasting: principles and practice (OTexts, 2021) – interesting book, full text online, code in R