find.neighbors: find.neighbors

In insilico/STIR:

Description

Find the nearest hit/miss matrices

Usage

find.neighbors(attr.mat, pheno.class, metric = "manhattan",
  method = "multisurf", k = 0, sd.vec = NULL, sd.frac = 0.5)

Arguments

attr.mat	m x p matrix of m instances and p attributes
pheno.class	length m vector of binary class status (usually -1/1) find.neighbors
metric	for distance matrix between instances ("manhattan" or "euclidean")
method	neighborhood method ["multisurf" or "surf" (k=0) or "relieff" (specify k)]
k	number of constant nearest hits/misses for "relieff"
sd.frac	multiplier of the standard deviation of the distances when subtracting from average for SURF or multiSURF. The multiSURF default is sd.frac=0.5: mean - sd/2

Value

return variable (hitmiss.list) is a two-element: hitmiss.list[[1]] (hits) and hitmiss.list[[2]] (misses). Each list has two columns: $Ri_idx is the first column (instances) in both lists. The second column is $hit_idx (nearest hits for the first column instance) for list [[1]] and $miss_idx (nearest misses) for list [[2]].

Examples

#See vignette("STIRvignette")
RF.method = "multisurf"
metric <- "manhattan"
neighbor.idx.observed <- find.neighbors(predictors.mat, pheno.class, k = 0, method = RF.method)
results.list <- stir(predictors.mat, neighbor.idx.observed, k = k, metric = metric, method = RF.method)
t_sorted_multisurf <- results.list$STIR_T
t_sorted_multisurf$attribute <- rownames(t_sorted_multisurf)

insilico/STIR documentation built on May 15, 2019, 2:51 a.m.