Recent time series analysis needs to use Python statsmodels, but the installation process encountered a headache, Google, stackover all kinds of have not found a suitable solution, And there seems to be a lot of classmates in the Spit slot Windows Python installation scipy is a mess, so it is necessary to share, to help people avoid this hole.
In general, numpy and pandas are essential for scientific calculations in Python, and pandas also relies on numpy. If the PIP is installed, the direct pip install NumPy can be installed, but SciPy relies on NUMPY+MKL, so the direct installation NumPy no inter mkl is not used, the recommendation is from http://www.lfd.uci.edu/~ Download the corresponding. whl file in the gohlke/pythonlibs/, for example, my Python version is 3.4, 32 bits, download NUMPY-1.11.0+MKL-CP34-CP34M-WIN32.WHL to the directory where Python is installed/ Scripts. If wheel has already been installed with Pip, then in cmd:
Pip Install NUMPY-1.11.0+MKL-CP34-CP34M-WIN32.WHL
The installation was successful. Another way is to rename the. WHL suffix to zip, unzip, unzip the Numpy,numpy-1.11.0.dist-info two folders into Python/lib/site-packages.
With NumPy and MKL, you can then install SciPy and statsmodels in the same way, note that the installation sequence is numpy+mkl,scipy,statsmodels.
The first is in the Pycharm installed always error missing lapack and Blas, after checking lapack mannual finally CMake and MinGW can also be solved, but the process is too headache, so I think the above method is relatively simple ~
Transfer from-----http://blog.csdn.net/spirit_go/article/details/51567363
Install Numpy+mkl,scipy and statsmodels under Windows python