用神经网络在R中计算？

Question

all tuples in allClassifiers tuples are either 1 or 2 eg

    naiveBayesPrediction    knnPred5    knnPred10   dectreePrediction   logressionPrediction    correctClass
    1                       2                1           1                     1                          1
    1                       2                1           1                     1                          1
    1                       2                1           1                     1                          1
    1                       2                1           2                     1                          1

I trained the ensembler

ensembleModel <- neuralnet(correctClass ~ naiveBayesPrediction + knnPred5 + knnPred10 + dectreePrediction + logressionPrediction, data=allClassifiers[ensembleTrainSample,])

but I am trying to use it to predict with this:

compute(ensembleModel, allClassifiers[ensembleTestSample,])$net.result

but I get this error:

Error in neurons[[i]] %*% weights[[i]] : non-conformable arguments

my train and Test samples

ensembleTrainSample <- sample(nrow(allClassifiers), nrow(allClassifiers)*0.7)
ensembleTestSample <- (1:nrow(allClassifiers))[!(1:nrow(allClassifiers))%in% ensembleTrainSample]

Answer 1

Similar to your other question this is an error the stems from matrix multiplication. In essence, the following error:

Error in neurons[[i]] %*% weights[[i]] : non-conformable arguments

means that your matrices don't have dimensions that match for matrix multiplication. It's like trying to multiply a 4x4 matrix by a 10x10 matrix. it simply isn't going to work.

Now the reason you are getting this error is because you overlooked something in the documentation. If you look at ?compute you will see the following note about the covariate argument`:

covariate   a dataframe or matrix containing the variables 
            that had been used to train the neural network.

The key here is VARIABLES , not your entire dataset nor the classifier variable (you are trying to predict this!). Here is an example again with the infert dataset.

library(neuralnet)
data(infert)

# fit your neuralnet model
net.infert <- neuralnet(case~parity+induced+spontaneous, infert)

net.pred <- compute(net.infert, infert[,c("case","parity","induced","spontaneous")])

Error in neurons[[i]] %*% weights[[i]] : non-conformable arguments

But if I just include the variables I used to create the model it works without error.

# no error
net.pred <- compute(net.infert, infert[,c("parity","induced","spontaneous")])