from which node is the output generated and on which data does it show greater accuracy?