Skip to content

Latest commit

 

History

History
122 lines (93 loc) · 3.16 KB

automation-demo.md

File metadata and controls

122 lines (93 loc) · 3.16 KB
Raw

Forecast Code

gflu = read.csv("http://www.google.org/flutrends/about/data/flu/us/data.txt",skip=11)
y = ts(gflu$Massachusetts)
arima_model = arima(y, order = c(3, 0, 1))
forecast = predict(arima_model, n.ahead = 10)

jpeg("forecast_plot.jpg")
plot(y,type='l',ylab="Flu Index",lwd=2,xlim=c(540, 640),ylim=c(0, 4000))
lines(forecast$pred, col = 'blue', lwd=2)
dev.off()

predictions = data.frame(time = time(forecast$pred),
                         prediction = forecast$pred,
                         prediction = forecast$se)
write.csv(predictions, "predictions.csv", row.names = FALSE)
  • Save as forecast.R

Local Repetition

Mac/Linux

On OS X and Linux you use cron to automatically run commands on a schedule:

crontab -e
m h d m dow /home/ethan/forecast.R
26 2 * * * /home/ethan/forecast.R

To change or remove just run crontab -e again.

Windows

  • Start (Windows key)
  • Task Scheduler
  • Create Basic Task (follow defaults)
  • Program/script Rscript
  • Arguments /home/ethan/forecast.R
  • Show task in Active Tasks
  • To change or delete double click task
  • **Delete

Remote Repetition

  • Add forecast.R to GitHub
  • Now set up to run whenever anythin in the repo is changed and put the updated results into the repo

Connect GitHub to Travis for Continuous Integration/Deployment

  1. GitHub
  2. Make a new Repo w/README
  3. https://travis-ci.com & login
  4. Activate GitHub Apps -> Select Repo
  5. GitHub -> Setting -> Dev Settings -> PAT
  6. Add w/repo permissions (be careful with security)
  7. Copy token
  8. Travis -> More Options -> Settings
  9. Environmental Variable -> name: GITHUB_TOKEN value: paste

Give CI instructions on what to do

  • First tell it what packages we need
  • Create install.R
  • We don't need any so we'll install cowsay as an example
install.packages("cowsay")
  • Next we need to tell the system what to do
  • Do this using a yml file that provides metadata describing what we want
  • Create .travis.yml
language: r
cache: packages

install:
  - Rscript -e "print('skipping standard install')"
  
script:
  - Rscript forecast.R

deploy:
  provider: pages
  skip_cleanup: true
  github_token: $GITHUB_TOKEN 
  keep_history: true
  target_branch: master
  • Show code running on Travis
  • Add cron job -> branch: master timing: daily
  • Show image and forecasts in GitHub

Binder/Containers

  • Add runtime.txt to repo
r-2019-08-01