npdrLearner: npdrLearner

In insilico/npdr: Nearest-neighbor Projected-Distance Regression

npdrLearner

Description

Uses npdr neighborhoods to learn a nearest neighbor classification or regression model (latter not implemented but easy). Finds the nearest neighbors of test instances to a training dataset. Uses majority class of training neighbors for test prediction. Allows adaptive Relief neighborhoods or specify k. Regression would simply use the average value of the neighbor phenotypes. Uses functions npdrDistances2 and nearestNeighbors2.

Usage

npdrLearner(
  train.outcome,
  train.data,
  test.outcome,
  test.data,
  nbd.method = "relieff",
  nbd.metric = "manhattan",
  msurf.sd.frac = 0.5,
  dopar.nn = FALSE,
  knn = 0
)

Arguments

train.outcome	character name or length-m numeric outcome vector for train data
train.data	m x p matrix of m instances and p attributes of train data. May also include outcome vector but then outcome should be a name. Include attr names as colnames.
test.outcome	character name or length-m numeric outcome vector of test data
test.data	m x p matrix of m instances and p attributes for test data. May also include outcome vector but then outcome should be a name. Include attr names as colnames.
nbd.method	neighborhood method: 'multisurf' or 'surf' (no k) or 'relieff' (specify k). Used by nearestNeighbors2().
nbd.metric	used in npdrDistances2 for distance matrix between instances, default: 'manhattan' (numeric). Used by nearestNeighbors2().
msurf.sd.frac	multiplier of the standard deviation from the mean distances; subtracted from mean for SURF or multiSURF. The multiSURF default is 'msurf.sd.frac=0.5': mean - sd/2. Used by nearestNeighbors2().
dopar.nn	whether or not neighborhood is computed in parallel, default as FALSE.
knn	number of constant nearest hits/misses for 'relieff' (fixed-k). Used by nearestNeighbors2(). The default knn=0 means use the expected SURF theoretical 'k' with 'msurf.sd.frac' (0.5 by default)

Value

list: neighborhoods for each test instance, prediction for each test instance, accuracy on test set

Examples

test.results <- npdrLearner(
  train.outcome = "class", train.data = case.control.3sets$train,
  test.outcome = "class", test.data = case.control.3sets$validation,
  nbd.method = "relieff", nbd.metric = "manhattan",
  dopar.nn = FALSE, knn = 0
)
test.results$accuracy

insilico/npdr documentation built on July 6, 2023, 1:14 p.m.