Having preprocessed data using weka supervised cfssubseteval forward and backward selection, the decision tree model builds and makes prediction well, but naive Bayes do not recognize the same Arff file of preprocessed data with 11 nominal attributes and 74 instances