- task.type
character, either "classif" or "regr".
- nobs
integer, number of observations uniformly sampled from the full data set.
- nfactors
character, controls the number of factors (categorical features) to use.
Can be "low", "med", "high", or "full" (full corresponds to original data set).
- nnumericals
character, controls the number of numerical features to use.
Can be "low", "med", "high", or "full" (full corresponds to original data set).
- cardinality
character, controls the number of factor levels (categories)
for the categorical features. Can be "low", "med", "high" (high corresponds to original data set).
- data.seed
integer, this will be used via set.seed() to make the random subsampling reproducible.
Will not have an effect if all observations are used.
- cachedir
character. The cache directory, e.g., "oml.cache"
.
Default: "oml.cache"
.
- target
character "age" or "income_class". If target = age
, the
numerical varible age
is converted to a factor:
age<-as.factor(age<40)
- cache.only
logical. Only try to retrieve the object from cache.
Will result in error if the object is not found. Default is TRUE.