SQB (version 0.4)

SQBreg: SQBreg

Description

Do prediction using sequential bagging method with tree based learning algorithm

Usage

SQBreg(data.train, data.test, y, res, reps, cores, FunKDE, control,
  SQBalgorithm.1, SQBalgorithm.2, k, ncomp, nnet.size)

Arguments

data.train

Training dataset

data.test

Testing dataset

y

Numeric response variable

res

Resampling size. Could not be greater than the input data size.

reps

Replicates for the first bagging, default 100

cores

Use multi-cores, default one core, use cores='maxcores' for full use.

FunKDE

Kernel density estimate function. Use different kernel to fit, default logistic kernel.

control

Use in rpart package, rpart.control to tune the control

SQBalgorithm.1

Use for the initial training. Option: CART, lm(default), knnreg, nnet, PCR.

SQBalgorithm.2

Use for the last training. Option: CART, lm(default), knnreg, nnet, PCR.

k

The number of nearest neighbour used for knnreg

ncomp

The number of component used for PCR

nnet.size

The number of hidden layer and neuron for nnet

Value

Given testing set input, make a regression prediction

References

Breiman L., Friedman J. H., Olshen R. A., and Stone, C. J. (1984) Classification and Regression Trees.

Soleymani, M. and Lee S.M.S(2014). Sequential combination of weighted and nonparametric bagging for classification. Biometrika, 101, 2, pp. 491--498.

Efron, B. (1979). Bootstrap methods: Another lo ok at the jackknife. Ann. Statist., 7(1):1-26.

Examples

# NOT RUN {
data(hills, package="MASS")
rt.df <- hills[sample(nrow(hills)),]
data.train <- rt.df[1 : (length(rt.df[, 1]) - 1), ]
data.test <- rt.df[-(1 : (length(rt.df[, 1]) - 1)),]
fit <- SQBreg(data.train, data.test, reps = 30, y = "time")
fit

# }