Question

compare the performance of Weka and python when doing association analysis. who do the pest, why ?

Answer 1

Comparing the performance of Weka and python :

1.Python is much easier to play around with, try out ideas, etc. WEKA is very verbose. It might be more robust but it's not a language to easily try stuff out. It's an enterprise-y language, which can be sort of a cludge if you want to write some quick data-mangling script or glue together some ML algorithms.
2.Python has a very healthy ML ecosystem. There are a lot of libraries out there, much of the research is done in those languages, and you have very good tooling for data-exploration. They have interactive sessions (REPL).
3."The speed benefit" isn't actually that large. While current python implementations are rather slow, they allow to easily interface with plain C code. Often, the ML libraries people use are written in plain C under the hood, and are optimized for speed.

1.The first thing that can be said is that WEKA does not need a programming knowledge. WEKA is for people that are using ML for quick runs without needing to understand it.
2.It uses tons of memory and inefficient implementations.
3.You are restricted by what WEKA offers you. If you want to glue algorithms together / tune them / try out your own ideas in ways that the restrictive WEKA framework doesn't allow, you have to dig around WEKA's source code and work with Java, which is wordy and just not flexible enough.

So if you are a statistician (with no knowledge of programming languages) then you can use WEKA by its graphic interface. This is an advantage that makes it very useful for this category of users.