tune alpha and lambda of Tweedie model with elastic net

elasticTweedie.r

library(caret)
library(HDtweedie)

elasticTweedie = list(library="HDtweedie", type="Regression")

prm = data.frame(parameter=c("alpha", "lambda"),
    class=rep("numeric", 2),
    label=c("alpha", "lambda_vec"))

elasticTweedieGrid = function(x, y, len=NULL, search="grid") {
    # we only tune alpha and lambda
    library(HDtweedie)

    if (search == "grid") {
        out = expand.grid(alpha=seq(0, 1, 0.1), lambda=seq(0, 100, 10))
    } else {
        stop("Not implemented yet!")
    }

    return(out)
}

elasticTweedieFit = function(x, y, param, last, ...) {
    HDtweedie::HDtweedie(x=as.matrix(x), y=y,
        alpha=param$alpha, nlambda=1, lambda=c(param$lambda), eps=1e-10)
}

elasticTweediePred = function(modelFit, newdata, ...) {
    predict(modelFit, newdata)
}

elasticTweedieProb = function(modelFit, newdata, preProc = NULL, submodels = NULL) {
  
}

elasticTweedie$parameters = prm
elasticTweedie$grid = elasticTweedieGrid
elasticTweedie$fit = elasticTweedieFit
elasticTweedie$predict = elasticTweediePred
elasticTweedie$prob = elasticTweedieProb

fitControl = trainControl(method="cv", number=3)
tuningGrid = expand.grid(alpha=seq(0,1,0.2), lambda=seq(0,100,20))
elasticTweedieModel = train(x=raw_data[,colnames(raw_data)!="y"], y=raw_data$y, method=elasticTweedie, trControl=fitControl, tuneGrid=tuningGrid)

fm = elasticTweedieModel$finalModel
print(fm$b0)  # get intercept
print(fm$beta)  # get coefficients