rm(list = ls(all = TRUE))


library(glmnet)
library(caret)

timeControl <- trainControl(method = "timeslice",
                            initialWindow = 950,
                            horizon = 1,
                            fixedWindow = FALSE,
                            returnData=TRUE,
                            returnResamp="all",
                            savePredictions=TRUE,
)


# create grid for parameters evaluation, alpha=1 -> LASSO

grid <- expand.grid(.alpha=seq(0,1,.1),
                     .lambda=seq(50,0,-.1),
)

# train the elnet model

trainModel <- train(y=seq(0.01,10,.01), x=matrix(rep(log(seq(.01,10,.01)),10), nrow = 1000),
                     method="glmnet",
                     metric="RMSE",
                     trControl=timeControl,
                     tuneGrid=grid,
)

> trainModel$pred[ which(trainModel$pred[,"lambda"]==0 & trainModel$pred[,"Resample"]=="Training01"),]
         pred  obs rowIndex alpha lambda   Resample
501  7.159336 9.51      951   0.0      0 Training01
1002 7.163996 9.51      951   0.1      0 Training01
1503 7.163973 9.51      951   0.2      0 Training01
2004 7.164281 9.51      951   0.3      0 Training01
2505 7.163766 9.51      951   0.4      0 Training01
3006 7.164378 9.51      951   0.5      0 Training01
3507 7.163292 9.51      951   0.6      0 Training01
4008 7.163606 9.51      951   0.7      0 Training01
4509 7.163839 9.51      951   0.8      0 Training01
5010 7.164014 9.51      951   0.9      0 Training01
5511 7.164132 9.51      951   1.0      0 Training01