R/d1.model.R

In overhuman/d1r: d1r package

d1.model.horseRace <- function(x,y,training,testSet){


	h2o.init()
	training = as.h2o(training)
	testSet = as.h2o(testSet)

	#h2o.gbm
	gbm=list()
	gbm$model= h2o.gbm(y=y,
					   x=x,
					   training_frame=training,
					   validation_frame=testSet)	
	preds.gbm = h2o.predict(gbm$model,testSet)
	actual = eval(parse(text=paste("testSet",y,sep="")))
	predicted = as.vector(preds.gbm)
	gbm$cor = d1.stat.cor(predicted,actual) 
	gbm$mse = d1.stat.MSE(actual,preds.gbm)
	gbm$hit = d1.stat.hit(predicted,actual)
	gbm$huber = d1.stat.huber(predicted,actual)
	gbm$rwMse = d1.stat.MSE(actual,0* preds.gbm *0) 
	gbm$rwHuber = d1.stat.huber(0* predicted *0,actual)
	
	
	#DL
	dl=list()
	dl$model= h2o.deeplearning(y=y,
							   x=x,
							   training_frame=training,
							   validation_frame=testSet)	
	preds.dl = h2o.predict(dl$model,testSet)
	actual = eval(parse(text=paste("testSet",y,sep="")))
	predicted = as.vector(preds.dl)
	dl$cor = d1.stat.cor(predicted,actual) 
	dl$mse = d1.stat.MSE(actual,preds.dl)
	dl$hit = d1.stat.hit(predicted,actual)
	dl$huber = d1.stat.huber(predicted,actual) 
	dl$rwMse = d1.stat.MSE(actual,0* preds.dl *0) 
	dl$rwHuber = d1.stat.huber(0* predicted *0,actual)

	#bagged trees

	#chart MSE 

	#chart Cor

	#Chart Hit


	# Chart Backtest

	return(list(gbm = gbm,
				))


}