Naive Bayes Classification for categorical values

Hi, I have a dataset containing numerical and categorical data. I like to use Naive Bayes Classifier in the following link but it only confers with numerical values.

http://www.mathworks.com/matlabcentral/fileexchange/37737-naive-bayes-classifier/content//NaiveBayesClassifier.m

Does anyone have an idea about how to extend that code to categorical data as well?

Answer 1

[Sean de Wolski]

Convert the categoricals to doubles.

c = categorical({'red';'blue';'red'})
double(c)

[Ege]

That's a good idea but I am using tables. When I do that, I loose my column names

converted=double(tNonNumeric{:,:});
coverted_arr=array2table(converted);
merged_w_converted= [tNumeric coverted_arr];

What can I do about that?

[Ege]

And also I'm getting Subscripted assignment dimension mismatch. error in following part of the code from the link above

for i=1:num_of_classes
            for j=1: height(xi)
                xi=x((y==yu(i)),:); 
                mu(i,:)=mean(xi{j,:},1); %it was originally mean(xi,1)
                sigma(i,:)=std(xi{j,:},1);
            end
        end

[Sean de Wolski]

The table is nice for storing but not for these computations. Reconvert to a table after with the same names from the original table (T.properties.variableNames).

[Ege]

Hi i just realized that converting categorical to double makes things complicated. For example it gives different values to the ones that are actually same. I see values 127 ,96 all sort of different ones in a column where all of them were actually 'yes'. So what should I do about it?