makeModel: make a mlr model + its metadata for a specific task using a...

In pokyah/agrometeoR: Spatial interpolation for Agromet project @ CRA-W

Description

make a mlr model + its metadata for a specific task using a specific learner

Usage

makeModel(task, learner)

Arguments

task	an object of class mlr::makeRegrTask()
learner	an object of class mlr::makeLearner()

Value

A 2 elements named list

• snitch : a boolean. Is TRUE if function has provided the expected result. Is FALSE is function throws an error

• output : a named list which elements are :

  • value : a named list which elements are :

    • trained : an element of class mlr::makeWrappedMode()

    • predictions an element of class data.frame containing the predictions made by the trained model at station locations. Colnames are truth (= observations), response (= predictions), se (= standard error) and residuals (= truth - response)

    • perfs : a named list which elements are :

      • iters : a dataframe containing the information relative to the leave-one-out cross validation of the model. Colnames are learner.id, iter, rmse, mse, mae, timetrain

      • agg : a dataframe containing the aggregated performances of the LOOCV. Same colnames as iters

  • condition : a character specifying the condition encountered by the function : success, warning, or error.

  • message : a character specifying the message relative to the condition.

Author(s)

Thomas Goossens

Examples

## Not run: 
# load magrittr for pipe use : %>%
library(magrittr)
# create the dataset
myDataset = makeDataset(
  dfrom = "2017-03-04T15:00:00Z",
  dto = "2017-03-04T18:00:00Z",
  sensor = "tsa")

# extract the list of hourly sets of records
myDataset = myDataset$output$value

# create the tasks
myTasks = purrr::map(myDataset, makeTask, target = "tsa")

# extract the tasks from the outputs
myTasks = myTasks %>% purrr::modify_depth(1, ~.$"output"$"value"$"task")

# keep the first task
myTask = myTasks[[1]]

# create the model
myModel = makeModel(
  task = myTask,
  learner = agrometeorLearners$mulLR_lonLatAlt_NA)

# extract the relevant information
myModel = myModel$output$value

# show the results of the LOOCV of this hour forthe choosen learner
head(myModel$predictions)
head(myModel$perfs$iters)

# show the model information
myModel$trained

## End(Not run)

pokyah/agrometeoR documentation built on May 26, 2019, 7 p.m.