I'm trying to create a NB classifier via RWeka and it is returning a variable type error.
I have the following variables:
dtm_df.train as a data.frame containing the following
ask check state
1_10 0 1 bad
1_100 1 0 bad
1_11 2 1 good
1_13 0 0 bad
1_14 0 0 good
1_15 0 1 bad
1_16 0 1 good
1_17 0 0 bad
1_19 0 0 bad
1_2 2 0 bad
and class.formula as a formula containing: state ~ ask + check
When using
NB <- make_Weka_classifier("weka/classifiers/bayes/NaiveBayes")
classifier <- NB(class.formula ~ ., dtm_df.train)
It returns:
Error in model.frame.default(formula = class.formula ~ ., data = dtm_df.train) : object is not a matrix
Converting the data parameter dtm_df.train to matrix does not work, since it requires a data.frame.
When trying:
classifier <- NB(class.formula ~ ., dtm_df.train)
It returns
Error in .jcall(o, "Ljava/lang/Class;", "getClass") :
weka.core.UnsupportedAttributeTypeException: weka.classifiers.bayes.NaiveBayes: Cannot handle string class!
Found out that the cause was actually the format of the state column in the train set dtm_df.train.
The solution was converting that column to factor via:
dtm_df.train$state <- as.factor(dtm_df.train$state)