Everything is difficult at the beginning, as the first blog, learn not to be easy to understand, fun, but to honestly make things clear.
The thing that originated from the Kaggle competition was generously opening the source on GitHub, and Kaggle was very thoughtful to sort out these excellent solutions and implementations. For small white-level data workers, such as me, is a perfect opportunity to copy ideas and learn code. In order to enjoy this feast, I built a python environment under Windows. Because ML packet dependence is a bit complex, this article is described in one or two.
Common ml of Python lib are: NumPy, Matplotlib, SciPy, Scikit-learn.
Way One:
A common tool for installing Python third-party packages under Windows Setuptools is small and handy. It omits the need to install Python third-party package download, decompression, execute Python **.py and other tedious process, default to Http://pypi.python.org/simple search third-party package and download the installation, similar to the Yun function under Linux.
As with Yun, it also checks for dependencies, so Numpy,matplotlib,scikit-learn will get an error during installation.
Way two:
I choose to install manually: SourceForge provides EXE installation package, seemingly do not rely on check, the above packages can be installed directly. There is an online view that Scikit-learn will rely on nose, which has not been confirmed. I easy_install nose first, then download the three packages and install them.
The trouble with this is that when you run Python code, these third-party packages are prompted to import * * *. Easy_install can be basically solved, such as six. Dateutil is an exception, and SourceForge does not provide an installation package.
Way three:
At this point, move to dateutil:http://www.lfd.uci.edu/~gohlke/pythonlibs/#python-dateutil
Pyparsing is also: http://www.lfd.uci.edu/~gohlke/pythonlibs/#pyparsing
The above three ways can basically get the required third party package.
Installation of the Windows ML Python lib