Hi, This mail is in regards with a query we have in usage of Cubist package in R. We are using the CRAN package - "Cubist" for predicting sales. We have a large training dataset which in turn gives a huge specification result file. (Note: Independent variables consists of both character and numerical variables) We use this specification file to predict for our testing data which has similar structure like training data used. The problem is when we provide large testing dataset at one go in the predict function of the Cubist package we get "0" as predicted value. But when we provide the testing dataset in small chunks into the predict function of the Cubist package we get appropriate values as the predictions. In order to do further research in this, we tried different combinations of data to identify the exact issue but we couldn't really identify the cause of the issue. In our opinion this problem is likely because of some memory/stack overflow exception in the Cubist package. A combination of number of rows and the complexity of operation for these rows (number of if conditions they satisfy in the specifications) might be causing this issue. Could you please clarify the issue ? Any help would be great. Thanks, Swarna Gupta [[alternative HTML version deleted]]