Since I got the chance to access to both Google Storage for Developers and Google Prediction API (more details
here and
here), I decided to create a simple wrapper (just 4 basic functions until now) to be capable to play with the Google Prediction API from R.
Here you can find the github repository for the project and below few lines of code reproducing an
example you can find on the Google Prediction API website.
Download the source code from
here.
Either source the functions contained in the R directory or install the package typing (from the command line in a Unix-like environment):
R CMD INSTALL predictionapirwrapper_1.0.tar.gz
# start R and type (code highlighting thanks to Revolution Analytics
Pretty R syntax highlighter):
library(predictionapirwrapper)
## The first stage of using the API is to acquire an authorization token. This can be done via this command:
token <- GetAuthToken(email="user@gmail.com", passwd="mypassword")
## This command begins training on data that has been previously uploaded to Google Storage.
GoogleTrain(auth_token=token$Auth, mybucket="data_languages", mydata="language_id.txt")
## Once training has started, this command checks the status of the training job and gets meta-information on the model (if available).
GoogleTrainCheck(auth_token=token$Auth, mybucket="data_languages", mydata="language_id.txt")
## When training has finished, this command issues a request for a new prediction from the model.
GooglePredict(auth_token=token$Auth, mybucket="data_languages", mydata="language_id.txt", myinput="La idioma mas fina")
All comments, corrections, alternative code are more than welcome!
Update: a more complete and functional alternative can be found
here.
