Great post!

I'm curious to know why did you scale the testing data with the information in the testset?

I believe you'd want to have the training and testing data both going through the same transformations, so the testing is not influenced by outside/ seen information.