Plots a variation of the squared prediction error (SPE) statistic to visualize the contribution of each variable to a fault.
mspSPEPlot(
trainData,
trainLabel,
trainSPE,
newData,
newLabel,
newSPE,
trainObs,
var.amnt
)an xts data matrix containing the training observations
Class labels for the training data as a logical (two states only) or finite numeric (two or more states) vector or matrix column (not from a data frame) with length equal to the number of rows in ``data." For data with only one state, this will be a vector of 1s.
the SPE values corresponding to the newLabel state calculated by mspTrain using the full training data with all variables included
an xts data matrix containing the new observation
the class label for the new observation
the SPE value returned by mspMonitor using the full new observation with all variables included
the number of observations upon which to train the algorithm. This will be split based on class information by a priori class membership proportions.
the energy proportion to preserve in the projection, which dictates the number of principal components to keep
if (FALSE) {
# Create some data
dataA1 <- mspProcessData(faults = "B1")
traindataA1 <- dataA1[1:8567,]
# Train on the data that should be in control
trainResults <- mspTrain(traindataA1[,-1], traindataA1[,1], trainObs = 4320)
# Lag an out of control observation
testdataA1 <- dataA1[8567:8568,-1]
testdataA1 <- lag.xts(testdataA1,0:1)
testdataA1 <- testdataA1[-1,]
testdataA1 <- cbind(dataA1[8568,1],testdataA1)
# Monitor this observation
monitorResults <- mspMonitor(observations = testdataA1[,-1],
labelVector = testdataA1[,1],
trainingSummary = trainResults$TrainingSpecs)
tD <- traindataA1[,-1]
tL <- traindataA1[,1]
nD <- testdataA1[,-1]
nL <- testdataA1[,1]
tO <- trainObs
vA <- 0.95
nSPE <- monitorResults$SPE
tSPE <- trainResults$TrainingSpecs[[nL]]$SPE
mspSPEPlot(tD,tL,tSPE,nD,nL,nSPE,tO,vA)
}