The Pricing Engine is price-aware forecasting used for decision making.
Python requirements: We use Python 3.5. See tools/scripts/install_packages.bat
for a tool to install the specific package versions required and for which usages they are required.
Please also see this package's License and the Microsoft's Privacy & Cookies statement
What follows are instructions how to install on our "main" environment. See below for an additionally supported AML Workbench environment. The Pricing Engine requires a Python==3.5
and the package Scipy==0.19.1
. If you do not already have Scipy installed, it can be difficult on Windows. On Anaconda distributions you can install it easily:
conda install Scipy==0.19.1
You will additionally need numpy==1.13.1
, pandas==0.20.3
, scikit-learn==0.19.0
, and statsmodels== 0.8.0
. Then you should be able to install the Pricing Engine via:
Optional packages: matplotlib
(graphs), cvxopt==1.1.9
(for Synthetic Control features), jupyter==1.0.0
(for the notebook examples)
We have done limited testing to ensure that the AzureML Workbench environment works as well: numpy==1.11.3
, pandas==0.19.2
, scipy==0.18.1
, scikit-learn==0.18.1
(you will need to install statsmodels==0.8.0
).
Note: Newer versions of packages may work but given our currently limited resources they are currently not tested or supported (contact us if you have specific environment specifications).
We have only tested the package on Windows 10 and Windows Server 2012R2 environments.
You can install the wheel by downloading it from the pkgrepo
folder and using
pip install pricingengine-X.X.X-py3-none-any.whl
See the lib/examples
folder.
See documentation at https://bquistorff.github.io/pricingengine/index.html.