antahiap / dsr-time-series-03

Basic approaches for time series analysis in Python.

Introduction to time series forecasting

Prediction is very difficult, especially if it's about the future - Niels Bohr

One day course designed to give an introduction to the practice of time series forecasting.

What you will learn

00

• what determines a good time series forecast
• applications of time series
• what trend, seasonality, stationarity are
• understand the simple baseline approaches
• understand log & difference transforms

01

• missing values

02

• know how to decompose a time series
• know how to check if our time series is a random walk
• understand what autocorrelation is

03

• options for engineering time features

04

• understand the different error metrics options

05, 06

• understand the issues in evaluating the out of sample performance in time series
• use XGBoost for time series modelling

07 - 10

• classical time series with the Holt-Winters method
• classical time series with ARIMA, SARIMA & SARIMAX

11

• exercise applying SARIMA

Further reading

Forecasting: Principles and Practice - Rob J Hyndman and George Athanasopoulos - textbook