posit::conf(2023) Workshop: Tidy time series and forecasting in R
Register now: http://pos.it/conf Instructor: Rob J Hyndman Workshop Duration: 2-Day Workshop This course is for you if you: • already use the tidyverse packages in R such as dplyr, tidyr, tibble and ggplot2 • need to analyze large collections of related time series • would like to learn how to use some tidy tools for time series analysis including visualization, decomposition and forecasting It is common for organizations to collect huge amounts of data over time, and existing time series analysis tools are not always suitable to handle the scale, frequency and structure of the data collected. In this workshop, we will look at some packages and methods that have been developed to handle the analysis of large collections of time series. On day 1, we will look at the tsibble data structure for flexibly managing collections of related time series. We will look at how to do data wrangling, data visualizations and exploratory data analysis. We will explore feature-based methods to explore time series data in high dimensions. A similar feature-based approach can be used to identify anomalous time series within a collection of time series, or to cluster or classify time series. Primary packages for day 1 will be tsibble, lubridate and feasts (along with the tidyverse of course). Day 2 will be about forecasting. We will look at some classical time series models and how they are automated in the fable package, and we will explore the creation of ensemble forecasts and hybrid forecasts. Best practices for evaluating forecast accuracy will also be covered. Finally, we will look at forecast reconciliation, allowing millions of time series to be forecast in a relatively short time while accounting for constraints on how the series are related
image: thumbnail.jpg
Transcript#
This transcript was generated automatically and may contain errors.
Hi, I'm Rob Hindman, I'm a professor of statistics at Monash University in Australia. I'll be at the POSIT conference this year leading a workshop on tidy time series and forecasting using R. It's relatively common for organisations to collect huge amounts of data over time and then find that their existing time series analysis tools are not always suitable to handle the scale, the frequency and the structure of the data that have been collected.
For example, you might collect hourly data, but the regular classes in base R aren't designed for that. Or you might have a lot of time series, but again base R is really designed for single series or small numbers of series. You might have time series stored in a data frame or a tibble, but then the modelling tools aren't set up for time series models.
In this workshop we're going to look at some packages and methods that have been developed to handle the analysis of large collections of time series in a tidyverse style.
In this workshop we're going to look at some packages and methods that have been developed to handle the analysis of large collections of time series in a tidyverse style.
Day one: time series analysis
It'll be a two-day workshop and on day one we will talk about time series analysis using these tools. We'll look at the tibble data structure for flexibly managing collections of related time series. We'll look at how to do data wrangling, data visualisations and exploratory data analysis.
We'll explore feature-based methods to explore time series data in high dimensions. A feature-based approach can be used to identify anomalous time series within a collection of time series, or to cluster or classify time series.
Primary packages for day one will be the tibble package, lubridate and feasts, along with the tidyverse of course.
Day two: forecasting
Day two will be about forecasting. We'll look at some classical time series models and how they are automated in the fable package and we'll look at ensemble forecasts and hybrid forecasts, and also best practices for evaluating forecast accuracy.
Finally we'll look at forecast reconciliation, allowing millions of time series to be forecast in a relatively short time while accounting for constraints on how the series are related to each other.
