goscorer

A HTTP scorer in GoLang using GIN and PMML files

This package illustrates the use of Golang to provide scores using a random forest.

The way it works:

Offline, the IRIS dataset was used to create a random forest in R and is then exported to PMML and stored under the `/static` path. The PMML is provided. The Golang HTTP server preloads the model when the service starts (command: `go run main.go`), Alternative, compile and start the server with Heroku: "heroku local web"

go-Gin receives the GET request, parses it, and returns a score (probability to is a Setosa)

Try it out:

curl --location --request GET 'http://localhost:8085/testing' \
--header 'Content-Type: text/plain' \
--data-raw '{
"sepal_length_cm": 5.1
,"sepal_width_cm": 3.5
,"petal_length_cm":5
,"petal_width_cm": 0.2
}'

should return 0.55

Generating PMML files:

You can use generate_pmml.R to generate PMML objects. The output is stored in static.