I am using SVM in matlab, and this error appears
"Error using svmclassify (line 149)
Some observations in TEST could not be classified. This is probably caused by NaN values in the data. Logical response can't contain missing values. You can convert the logical response into NOMINAL or double."
can anyone help me to solve this problem?
I don't have a solution but a workaround. If the attribute is categorical I use most frequently seen items instead of missing values. If it's numerical I use mean of the available values.
Assume that TRNX and TSTX are input matrices for training and testing cases respectively. Each row represents a sample and each column represents a feature.
feature_count = size(TRNX,2);
for i = 1:feature_count
R = isnan(TRNX(:,i)); %index of missing values for training set
S = isnan(TSTX(:,i)); % index of missing values for testing set
if isCategorical(i)
m = mode(TRNX(~R,i)); % mode retrieves mostly seen value
else
m = mean(TRNX(~R,i)); % mean of available values
end
%assign m instead of missing values
TRNX(R,i)= m;
TSTX(S,i)= m;
end
hi Omar,
At any machine learning data you will find a file explain the nature of the data and how much missing attributes which the matlab deal with this missing data as NAN
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training