So you’ve spent a big 70% of your time munging the data together, another 10% of the time cleaning / normalising / transforming data, another 10% of the time finding the best set of features and the remaining 10% ensuring that it works well on the test set.
So after all that, you still have to go and put your model into production!
Maybe production is within an Excel spreadsheet, maybe its embedded in a Java application or maybe its embedded in a mobile application. It may be even using R in production in which case it will most likely take sys-admin resource, server resource and R-integration time to get the process running where it needs to be.
Whatever the case is, its going to take some effort to port your nice R code to the new production environment.
One way around all of this is to deploy your model to the cloud using DukeDeploy.
All that is required is
- – Sign-up for an API key
- – Build your model in R
- – Install the ddeploy R package > install.packages(“ddeploy”)
- – As per package examples, deploy to the cloud.
Once your model is deployed, its available for predictions immediately. Login to the accounts page at anytime to see your account usage details. There are some API limitations, see the documentation for more details etc.
More concrete example use-case posts to follow …