currently i have some trouble and question zu implement some kind of special missing value handling in WEKA J48 algorithm using WEKA JAVA API.

I want to test the performance of SHAPIRO approach zu handle missing value with imputation by using a decision tree to predict the value. Therefore, a separate decision tree for each attribute is needed. For a better understanding of my apporach, i had made a pseudo code: http://s7.directupload.net/images/140725/r4492x8v.png

My first problem: When i want to use the training data and set the class index to a continuous attribute, because of some missing values in, J48 cant work with continuous classes. So i have to discretize first. How can i discretize the training data in java code using WEKA API? The second problem: How can i change the class index of the training data after reading it? I only know how to do that during reading. Because i have du switch them to all attributes with missing in it and build decision trees.

Thanks for your help!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.