machine learning - The Role of the Training & Tests Sets in Building a Decision Tree and Using it to Classify -
i've been working weka couple of months now. currently, i'm working on machine learning course here in ostfold university college. need better way construct decision tree based on separated training , test sets. come idea can of great relief. thanx in advance.
-neo
you might asking more specific, in general:
you build decision tree training set, , evaluate performance of tree using test set. in other words, on test data, call function named c*lassify*, passing in newly-built tree , data point (within test set) wish classify.
this function returns leaf (terminal) node tree data point belongs--and assuming contents of leaf homogeneous (populated data single class, not mixture) have in essence assigned class label data point. when compare class label assigned tree data point's actual class label, , repeat instances in test set, have metric evaluate performance of tree.
a rule of thumb: shuffle data, assign 90% training set , other 10% test set.
Comments
Post a Comment