I do not track what you are after here. It makes no sense to convert coordinates to a "categorical" variable in regard to "feature means" in a recursive partitioning model. However, that aside, it is sensible to convert geographic coordinates to a distance based projection (projection units in meters or feet). In this way, the scaled statistical relationships are in distance and not time. Adding spatial coordinates to a model is referred to as a providing "naive spatial structure".
You can coerce your data into a sp class object and then use spTransform to convert the coordinates to something more relevant. Since you do not provide much information in your question it is difficult to provide any additional guidance. Here is a quick example of converting your data to a projected system.
dat <- data.frame(latitude = c(41.775974, 42.184913, 41.682957373),
longitude = c(-71.329887,-71.9179,-71.56037000))
coordinates(dat) <- ~longitude+latitude
proj4string(dat) <- "+proj=longlat +ellps=GRS80"
dat <- spTransform(dat, CRS=CRS("+proj=merc +ellps=GRS80"))
( dat.coords <- coordinates(dat) )