tests/testthat/test_multi_model_1.R

In manymodelr: Build and Tune Several Models

test_that(desc="Test that multi_model_1 works as intended",
                    code={
                      
                      skip_on_oldrel()
                      set.seed(520)
                      train_set<-createDataPartition(yields$normal,
                                                     p=0.8,
                                                     list=FALSE)
                      valid_set<-yields[-train_set,]
                      train_set<-yields[train_set,]
                      ctrl<-trainControl(method="cv",
                                         number=5)
                      m<-multi_model_1(train_set,"normal",".",
                                       c("knn","rpart"), 
                                       "Accuracy",ctrl,
                                       new_data =valid_set)
                       
                expect_error(multi_model_1(yields[1:120,],"normal",
                                           ".",c("knn","svmRadial"),
                                                     "Accuracy",ctrl),
              "new_data,metric,method, and control must all be supplied",
              fixed=TRUE)
              
             
            # need both control and metric
            expect_error(multi_model_1(yields[1:120,],
                                       "normal",".",
                                       c("knn","svmRadial"),
                                                 metric=NULL,ctrl,
                                       new_data = yields[1:120,]),
                    "new_data,metric,method, and control must all be supplied",
                                   fixed=TRUE)
            # need method
            expect_error(multi_model_1(yields[1:120,],
                                       "normal",".",method=NULL,
                                  metric="Accuracy",ctrl,
                                  new_data = yields[1:120,]),
                                   "new_data,metric,method, and control must all be supplied",
                                   fixed=TRUE)
            # ensure that no value is null
            expect_false(any(is.null(m$metric),is.null(m$predictions)))
                    })