All Hyndsight posts by date
library(tidyverse) library(tsibble) library(lubridate) library(feasts) library(fable) In my previous post about the new fable package, we saw how fable can produce forecast distributions, not just point forecasts. All my examples used Gaussian (normal) distributions, so in this post I want to show how non-Gaussian forecasting can be done.
As an example, we will use eating-out expenditure in my home state of Victoria.
vic_cafe <- tsibbledata::aus_retail %>% filter( State == "Victoria", Industry == "Cafes, restaurants and catering services" ) %>% select(Month, Turnover) vic_cafe %>% autoplot(Turnover) + ggtitle("Monthly turnover of Victorian cafes") Forecasting with transformations Clearly the variance is increasing with the level of the series, so we will consider modelling a Box-Cox transformation of the data.
The fable package for doing tidy forecasting in R is now on CRAN. Like tsibble and feasts, it is also part of the tidyverts family of packages for analysing, modelling and forecasting many related time series (stored as tsibbles).
For a brief introduction to tsibbles, see this post from last month.
Here we will forecast Australian tourism data by state/region and purpose. This data is stored in the tourism tsibble where Trips contains domestic visitor nights in thousands.
In my last post, I showed how the feasts package can be used to produce various time series graphics.
The feasts package also includes functions for computing FEatures And Statistics from Time Series (hence the name). In this post I will give three examples of how these might be used.
library(tidyverse) library(tsibble) library(feasts) Exploring Australian tourism data I used this example in my talk at useR!2019 in Toulouse, and it is also the basis of a vignette in the package, and a recent blog post by Mitchell O’Hara-Wild.
This is the second post on the new tidyverts packages for tidy time series analysis. The previous post is here.
For users migrating from the forecast package, it might be useful to see how to get similar graphics to those they are used to. The forecast package is built for ts objects, while the feasts package provides features, statistics and graphics for tsibbles. (See my first post for a description of tsibbles.
There is a new suite of packages for tidy time series analysis, that integrates easily into the tidyverse way of working. We call these the tidyverts packages, and they are available at tidyverts.org. Much of the work on these packages has been done by Earo Wang and Mitchell O’Hara-Wild.
The first of the packages to make it to CRAN was tsibble, providing the data infrastructure for tidy temporal data with wrangling tools.
One of the few people in Australia who did not write off a possible Coalition win at the recent federal election was Peter Ellis. We’ve invited him to come and give a talk about making sense of opinion polls and the Australian federal election on Friday this week at Monash University. Visitors are welcome. Here are the details.
11am, 31 May 2019. Room G03, Learning and Teaching Building, 19 Ancora Imparo Way, Clayton Campus, Monash University
I’ve tried my hand at writing for the wider public with an article for The Conversation based on my paper with Di Cook and Jeremy Forbes on “Spatial modelling of the two-party preferred vote in Australian federal elections: 2001-2016”. With the next Australian election taking place tomorrow, we thought it was timely to put out a publicly accessible version of our analysis.
There are now translations of my forecasting textbook (coauthored with George Athanasopoulos) into Chinese and Korean.
The Chinese translation was produced by a team led by Professor Yanfei Kang (Beihang University) and Professor Feng Li (Central University of Finance and Economics). The following students were also involved: Cheng Fan, Liu Yu, Long Xiaoyu, Wang Xiaoqian, Zeng Jiayue, Zhang Bohan, and Zhu Shuaidong.
The Korean translation was produced by Dr Daniel Young Ho Kim.
We currently have two postdoc opportunities together with an industry partner in the field of wind and solar power forecasting (full time, Level B). They are suitable for recently graduated PhD students that can start between now and June-July.
The opportunities are as follows:
Wind power forecasting: 1 year contract Good programming skills in R and/or Python Solid background in Machine Learning and/or Statistics Background in time series forecasting desirable Solar power forecasting: 6 months contract Good programming skills in R and/or Python Solid background in Machine Learning and/or Statistics Data will be cloud coverage data from sky cams, so some image processing background is necessary Background in time series forecasting desirable Please contact Christoph Bergmeir if you are interested.
For students who are interested in doing a PhD at Monash under my supervision.
First, check that you satisfy the following criteria:
You must have completed a degree in statistics that involved some research component (e.g., an honours or masters thesis). A degree in computer science, mathematics or econometrics might be acceptable if it contained a substantial amount of statistics. A degree in any other field is not sufficient background to work with me.