Am 03.03.2020 um 20:50 schrieb Artur Tarassow:
let me make some advertisement for a hobby project I've realized
last
weekend. The John Hopkins University publishes public data on world-wide
contagion of the virus. So I thought "Let's make use of it!"
Hi Artur, interesting idea. Let's just not forget the real suffering
involved in the background. (I know you didn't, but it may be worth
making it explicit when using words like "hobby" and so on.)
I've build a small job which automatically downloads a csv
(scheduled
via a cron job) and transforms data into a panel. Next, the job computes
7-days ahead forecasts based on latest available daily data for various
country-province combination. Freshly compiled plots are uploaded to
github. The underlying model is of the simple ARIMA type -- nothing
fancy as it's mainly for for illustration on how to use gretl for such a
task.
Or you could use your "naive" forecasting package, in particular the
model combination. Probably hard to beat!
By the way, the job is running on a Raspberry 4 with 2GB RAM -- and
just
works nicely. It wasn't difficult to compile gretl on it and required
dependencies are almost the same as for an Ubuntu system.
Yeah, it's impressive what you get these days for 30 or 40€.
cheers
sven