rknnBeg {rknn} | R Documentation |
Recursive Backward Elimination Feature Selection with Random KNN
rknnBeg(data, y, k = 1, r = 500, mtry = trunc(sqrt(ncol(data))), fixed.partition = FALSE, pk = 0.5, stopat = 4, cluster=NULL, seed = NULL) rknnBel(data, y, k = 1, r = 500, mtry = trunc(sqrt(ncol(data))), fixed.partition = FALSE, d = 1, stopat = 4, cluster=NULL, seed = NULL)
data |
An n x p numeric design matrix. |
y |
A vector of responses. For a numeric vector, Random Knn regression is performed. For a factor, Random classification is performed. |
k |
An integer for the number of nearest neighbors. |
r |
An integer for the number of base KNN models. |
mtry |
Number of features to be drawn for each KNN. |
fixed.partition |
Logical. Use fixed partition of dynamic partition of the data into training and testing subsets for each KNN. |
pk |
A real number between 0 and to indicate the proportion of the feature set to be kept in each step. |
d |
A integer to indicate the number of features to be dropped in each step. |
stopat |
an integer for the minimum number of variables. |
cluster |
An object of class ‘c("SOCKcluster", "cluster")’ |
seed |
An integer seed. |
Shengqiao Li<lishengqiao@yahoo.com>