MATLAB: Systems of double class cannot be used with the “predict” command.??

idsspredictsvm

hello there… i am new to svm. i got error while using "predict" as….. "Systems of double class cannot be used with the "predict" command. Convert the system to an identified model first, such as by using the "idss" command." i tried "idss" and i am getting error .. "The number of inputs and outputs of the model must match that of the data." What exactly should be done to avoid this issue? kindly help me

Thanks in advance.

Best Answer

You are calling svmtrain to create svmstruct . That structure will be an SVMStruct, described by svmtrain() documentation

You are calling idss() passing in a single parameter, svmstruct . According to http://www.mathworks.com/help/ident/ref/idss.html#inputarg_sys0 when you call idss() with only one parameter, that parameter must be:

"Any dynamic system to convert to an idss model:

When sys0 is an identified model, its estimated parameter covariance is lost during conversion. If you want to translate the estimated parameter covariance during the conversion, use translatecov.
When sys0 is a numeric (non-identified) model, the state-space data of sys0 define the A, B, C, and D matrices of the converted model. The disturbance matrix K is fixed to zero. The NoiseVariance value defaults to eye(Ny), where Ny is the number of outputs of sys."

The description of Dynamic Models at http://www.mathworks.com/help/ident/ug/dynamic-system-models.html lists routines that can produce identified models. All of those routines start with the two characters 'id'. You do not call any routines that start with 'id' so you are not creating any identified model. The routines that produce numeric models are tf, zpk, ss, frd, pid, pidstd, pid2, and pidstd2. You are not calling any of those. We must conclude that svmstruct is not an appropriate input to use for idss if you are passing only one parameter.

Related Solutions

MATLAB: Does sequentialfs always outperform cross-validation with selected features

I don't know if your code is correct. But accuracy estimates obtained by sequential selection are always biased high.

Consider say 10 random variables. Suppose you wish to find the variable with the largest true mean. Suppose these random variables are identical. Generate a separate sample for each variable. Due to the finite sizes of the samples, their estimated means are not going to be equal. You then choose the sample with the largest average and believe that the respective variable has the largest true mean. But all you did was choose the variable whose estimated mean came out largest by chance. Since the estimated mean is largest, it is likely above the true mean. Then you generate another sample for the chosen variable. Because the true mean is less than the estimated mean, your new estimate is less than your previous estimate.

This is exactly why you need to re-estimate the accuracy by another run of cross-validation after selection is done.

MATLAB: Adding parameter and value pair in fitcdiscr for QDA classifier

Passing KFold to any classification model creates a ClassificationPartitionedModel, which is a set of multiple models. As such, one cannot call predict on it. Instead, you can use kfoldPredict function, which gives you the classification accuracy on the training partition for each of the model. For example,

load fisheriris.mat
mdl = fitcdiscr(meas, species, 'DiscrimType', 'quadratic', 'KFold', 4);
kfoldPredict(mdl)

If you want to predict on test dataset, train a separate model with same parameters, except for KFold, and then use predict on the test dataset.

The error message is a known issue, and might be fixed in any of the upcoming releases.

Best Answer

Related Solutions

MATLAB: Does sequentialfs always outperform cross-validation with selected features

MATLAB: Adding parameter and value pair in fitcdiscr for QDA classifier

Related Question