ExprsPipeline
ObjectpipeFilter
subsets an ExprsPipeline
object.
pipeFilter(object, colBy = "valid.acc", how = 0, gate = 0, top = 0)
An ExprsPipeline-class
object.
A character vector or string. Specifies column(s) to use when filtering by model performance. Listing multiple columns will result in a filter based on the product all listed columns.
A numeric scalar. Arguments between 0 and 1 will impose
a threshold or ceiling filter, respectively, based on the raw value of
colBy
. Arguments between 1 and 100 will impose a filter based on
the percentile of colBy
. The user may also provide "midrange",
"median", or "mean" as an argument for these filters.
A numeric scalar. Determines the top N models based on
colBy
to include after the threshold and ceiling filters.
In the case that the @summary
slot contains the column "boot",
this selects the top N models for each unique bootstrap.
An ExprsPipeline-class
object.
The filter process occurs in three steps. However, the user may skip
any one of these steps by setting the respective argument to 0
.
First, a threshold filter gets imposed. Any model with a performance
less than the threshold filter, how
, gets excluded. Second,
a ceiling filter gets imposed. Any model with a performance greater
than the ceiling filter, gate
, gets excluded. Third, an
arbitrary subset occurs. The top N models in the ExprsPipeline
object get selected based on the argument top
. However,
in the case that the @summary
slot contains the column "boot",
pipeFilter
selects the top N models per bootstrap.
pipeFilter
will apply this filter based on the performance
metrics listed in the colBy
argument. Listing multiple columns
will result in a filter based on the product of all listed columns.
To weigh one metric over another, list that column more times.