g3rv4/RandomForest

A tiny library to run in C# random forests created in R

RandomForest

A simple implementation of a random forest in C#.

What is this?

This is a tiny library that knows how to parse PMML random forests and build predictions from them. If you want to train a model, then this library is not for you and you may be looking for something more like Accord.NET.

How can I install it?

On the Package Manager Console, run

Install-Package RandomForest

Can I see an example?

I'm glad you asked. I built a sample project using it.

What use case does this solve?

You use R to build and tune your models. You may be using caret to help you build and choose the one you like the most... but then, you want to productionize the model in C#.

In your R code, you can do:

library(r2pmml)
r2pmml(my_model, "my_model.pmml")

And that will generate a my_model.pmml that you can feed this library using:

RegressionRandomForest randomForest; // If you have a classification random forest, then use the ClassificationRandomForest class instead
using (XmlReader reader = XmlReader.Create("my_model.pmml"))
{
    randomForest = new Forest(reader);
}

You can use that randomForest to make predictions that match the ones you get on R.

var row = new Dictionary<string, string> {
    ["Name"] = "Gervasio Marchand",
    ["Age"] = "34"
}

double predictedValue = randomForest.Predict(row);

A more efficient alternative (because it doesn't need to parse the string to double) receives a Dictionary<string, double>. The key of the dictionary should be the variable name and the value when the variable type is not a double. Here's an example:

var row = new Dictionary<string, double> {
    ["NameGervasio Marchand"] = 1,
    ["Age"] = 34
}

double predictedValue = randomForest.Predict(row);

And that's it! I'm playing with protobuf-net to make it serializable... if I can avoid the dependency, that'd be nice.