Optimizing ME

Computes optimal category orders for each dimension separately. Uses a TSP solver to achieve the best ME value.

optME(x, dims = NULL, nstart = 1,
 solver = "nearest_insertion",
  return.table = TRUE, adjust.dist = FALSE)

Arguments

x	A matrix, table or array.
dims	Which dimensions to reorder. The dimensions are reordered independently.
nstart	The number of different starting points for the TSP solver. If `nstart` is greater or equal to the number of cities in a dimension, the solver uses each city once.
solver	Should be one of `"nn", "nearest_insertion", "cheapest_insertion","farthest_insertion"`. See solve_TSP.
return.table	Whether or not to return the optimized table. If `FALSE` only the new category orders are returned. If `TRUE` the new orders are attached to the table as an attribute `"orders"`.
adjust.dist	If `TRUE` a small proportion of the euclidean distances between the category profiles (e.g. rows) is added to the ME distance value. The idea is to keep identical profiles together which is otherwise not guaranteed, since the ME values can be identical even if the profiles aren't.

Each dimension is optimized separately via a TSP solver.

The passed object as a table with optimized category orders.

a <- arsim(2000,c(8,9,10),3,0.2)
ME(a)
#> [1] 46222
a2<-optME(a)
ME(a2)
#> [1] 97428