Machine learning is hard! Spent a whole day with Weka and it's Neural Networks. God my brain. There is too much to know before being really equipped to use this tool... especially from code.

Throw weka away, it is very bad for actual ML work. use tensorflow, or Spark. Weka is good for research or small scale data sets.

ML has two parts - data domain, and ML algorithem. You need to know the quality of the data, and match the optimal algorithem (or algorithems) to it.

I agree with @magicMirror, tensorFlow will empower you much faster

I’ve been playing with scikit, but I’m hating python with a passion.

Nice language I’m sure, but the breaking changes and dynamic nature of it basically seems to make people think they have a free pass to switch things up all the time.

Most tutorial code fails to run because it’s all different now, and then having things break at runtime because of a code error really pisses me off.

Machine learning is hard. I’m terrible at maths, so I’m slowly exposing myself to the concepts and playing around with it all.

Use Keras (if you work with images mostly) or Pytorch (best for Nlp) , when you get more advanced use tensorflow. If you want to do anything with real big data, take a look at BigDL or Spark ML.

Thanks for all the suggestions. I certainly will try these out, once I've gotten myself out of this mess.

You should learn the underlying concepts before jumping to any of those super-high abstraction toolsets. Read a book or two of underlaying math and logic. You will thank yourself later.