What is the meaning of interaction depth in gbm using 'caret' in R




While using gbm for a classification problem I came upon the interaction.depth option in the tunGrid function for gbm using caret

gbmGrid <-  expand.grid(interaction.depth = c(1, 5, 9),
                        n.trees = (1:30)*50,
                        shrinkage = 0.1,
                        n.minobsinnode = 20)

I am not being able to understand what this option does and why is it used?
Can someone please help me with this??



you find good explanations in the following paper, jump over the technical and go to paragraph 3.2 and 3.3 this will give you good start. GBM explanation.
If you want to know more than that always refer to the book of Kuhn “Applied Predictive Modelling”, you will have good explanations about Caret as well and Max Kuhn is the main person behind Caret. Max wrote few tutorials as well, check on the web and of course the Caret pages, which include example with gym. Caret Pages
Hope this help.