[ANN] ForecastData: Time Series Datasets for Julia

ForecastData.lj has been created in the process of decoupling functionality from Forecast.jl .

This package handles access to small to medium size time series datasets (in the Megabytes) to be used in Julia.

Since the number of datasets could be potentially large and relatively big in size they are not downloaded when the package is installed, but only when they are called to then be stored locally as a Julia artifact.

The format chosen to return the time series is via a DataFrame with its first column using Dates when appropiate.

In general time series available in ForecastData will be “unique” in one of these two ways:

  • They are a combination of publicly available datasets (e.g. london dataset)
  • They are publicly available datasets with some preparation to facilitate their use (e.g. quakes dataset)

Simple publicly available datasets will not be generally considered, for instance R datasets like AirPassengers will not be stored in ForecastData since they can be easily accessed via RCall.

The current available datasets are:

Name Description
co2 Atmospheric Carbon Dioxide Dry Air Mole Fractions from quasi-continuous measurements at Mauna Loa, Hawaii from 1973 to 2020
london Monthly data on weather and crime in Greater London from 2008 to 2018
quakes Earthquakes with a magnitude higher or equal to six from 1638-06-11 to 2021-10-02
seaborne Seaborne trade for France, Germany and the United Kingdom from 2015-04-01 to 2021-05-02

The link is pointing towards ForecastPlots.jl

EDIT: Nice work in the whole Forecast*.jl package families.
I am using in a Time Series lecture on scientific computing and data science with Julia

1 Like

I fixed it, thanks!

wow! This made my day! Thank you for letting me know @Storopoli! :heart:

Actually, the idea behind the future Forecast.jl v.0.2.0 is to integrate all these decoupled packages (and possibly more) to create an analytical environment for time series analysis that would be friendly for introductory courses. After publishing v.0.2.0 I was planning on asking feedback if anybody (professors and students) was using it for lectures but there you are in version v.01.0! Although future version v0.2.0 will be quite different; hopefully better and friendlier, any suggestions to improve usability are very welcome!

By the way, since we are at it, soon I’ll be announcing a new package in these decoupling series work that might be useful for your lectures too: GitHub - viraltux/DataWrangler.jl: Data transformation tools for analytics , it’s nearly finished so you can have a go at it already.

1 Like

Here it is the Lecture. It is in Portuguese. It will be presented not this Friday, but next Friday. So if you have suggestions there is time to incoporate them.


1 Like