Phd Report 9-Using package “caret” for building models based on delay 7 hours(P7) database

In this part, base on P7 build and tune 4 models ( svm,nnet,randomForest,lm) sperately by using package “caret”,with leave-one-out crossvalidation method of trainControl which number is 5,repeats is 5

load("/home/gong/prepareData/P 7 TrainingAndTesting.RData")
library(doMC)
library(kernlab)
library(PerformanceAnalytics)
library(caret)
registerDoMC(cores = 3)
# svm RBF kernel
cvcontrol <- trainControl(method = "LOOCV", number = 5, repeats = 5)
if (file.exists("P7.svmFit.RData")) {
    load("P7.svmFit.RData")
} else {
    P7.svmFit <- train(InputsTrain, TargetTrain, method = "svmRadial", tuneLength = 4, 
        trControl = cvcontrol, scaled = TRUE)
    save(P7.svmFit, file = "P7.svmFit.RData")
}
# neural networks
if (file.exists("P7.nnetFit.RData")) {
    load("P7.nnetFit.RData")
} else {
    nnet.grid <- expand.grid(.size = c(7:15), .decay = c(1e-04, 2e-04, 0.005, 
        0.01))
    P7.nnetFit <- train(InputsTrain, TargetTrain, method = "nnet", trControl = cvcontrol, 
        tuneGrid = nnet.grid)
    save(P7.nnetFit, file = "P7.nnetFit.RData")
}
# random Forests
if (file.exists("P7.rfFit.RData")) {
    load("P7.rfFit.RData")
} else {
    library(randomForest)
    P7.rfFit <- train(InputsTrain, TargetTrain, method = "rf", trControl = cvcontrol, 
        tuneLength = 3)
    save(P7.rfFit, file = "P7.rfFit.RData")
}
# Linear Least Squares
if (file.exists("P7.lmFit.RData")) {
    load("P7.lmFit.RData")
} else {
    P7.lmFit <- train(InputsTrain, TargetTrain, method = "lm", trControl = cvcontrol, 
        tuneLength = 4)
    save(P7.lmFit, file = "P7.lmFit.RData")
}

In this part, I will make prediction according different models,plot and calculate the errors

# the function to caculate the model errors
modelErrors <- function(predicted, actual) {
    sal <- vector(mode = "numeric", length = 3)
    names(sal) <- c("MAE", "RMSE", "RELE")
    meanPredicted <- mean(predicted)
    meanActual <- mean(actual)
    sumPred <- sum((predicted - meanPredicted)^2)
    sumActual <- sum((actual - meanActual)^2)
    n <- length(actual)
    p3 <- vector(mode = "numeric", length = n)
    for (i in c(1:n)) {
        if (actual[i] == 0) {
            p3[i] <- abs(predicted[i])
        } else {
            p3[i] <- ((abs(predicted[i] - actual[i]))/actual[i])
        }
    }
    sal[1] <- mean(abs(predicted - actual))
    sal[2] <- sqrt(sum((predicted - actual)^2)/n)
    sal[3] <- mean(p3)
    sal
}
# Predicting different models and plot the prediction values and true
# values
models <- list(svm = P7.svmFit, nnet = P7.nnetFit, randomForest = P7.rfFit, 
    lm = P7.lmFit)
P7.preValues <- extractPrediction(models, testX = InputsTest, testY = TargetTest)
plotObsVsPred(P7.preValues)

# calculate errors
P7.error <- function(model) {
    pd <- predict(model, newdata = InputsTest)
    modelErrors(pd, TargetTest)
}
rf.error <- P7.error(P7.rfFit)
nnet.error <- P7.error(P7.nnetFit)
svm.error <- P7.error(P7.svmFit)
lm.error <- P7.error(P7.lmFit)
errorAll <- rbind(rf.error, nnet.error, svm.error, lm.error)
errorAll

##               MAE   RMSE   RELE
## rf.error   0.1236 0.1538 0.5972
## nnet.error 0.1231 0.1531 0.5746
## svm.error  0.1229 0.1569 0.5633
## lm.error   0.1307 0.1613 0.6366