I'm in the middle of taking the Coursera Data Science course which is currently discussing machine learning. The current assignment is to have a go at a Kaggle competition and write up the attempt. (Kaggle is a site where data modelling competitions are posted: participants are given some existing data set and the aim is to develop models to make predictions or decisions for unknown data.)

This is why, earlier today, I was trying to install scikit-learn, a Python machine learning module. Unfortunately, when using pip to install it, I was getting the unhelpful error: "unable to find vcvarsall.bat". This blog post pointed out the issue: you need a C++ compiler installed and configured correctly (e.g. MinGW or Visual Studio).

Instead of having to do more installation and setup, the quickest solution was to use a portable Python distribution, like WinPython or Anaconda; these have several scientific modules installed already. Another option to using an existing Python distribution is to just install a precompiled package: there's also Christoph Gohlke's collection of compiled Python scientific packages available.