Important notice, September 2020: This repository was last updated in July 2015, while the GPII's real-time framework, for which the Statistical Matchmaker provided a component, has evolved considerably since that time. Since compatibility with that framework cannot be guaranteed, this repository will be archived.
The analysis module is one of several components of the GPII/Cloud4all Statistical Matchmaker. The other components are the runtime component and the training data.
How do these three components fit together? First, the analysis component analyses all the available preference sets (cf. the training data) and outputs data structures that the runtime component can use in order to responde to matchmaking requests. Since the clustering algorithms used by the analysis component are computationally intensive, these algorithms would only run at certain regular intervals, e.g. once a day. Second, the runtime component uses the data generated by the analysis component to respond to matchmaking requests.
The current training data will be phased out when the number of real GPII/Cloud4all becomes big enough to use their preference sets for clustering.
- .NET Framework 4.5 (although the project also compiles when you set the target framework to .NET Framework 3.5)
- Training data
Simply run the following command in a directory of your choice:
git clone https://github.com/REMEXLabs/GPII-Statistical-Matchmaker-Analysis
On Microsoft Windows, open the project in Microsoft Visual Studio Express or another IDE in the Visual Studio family and press F7 ("Build Solution").
(You can probably also use SharpDevelop but we have not tested this.) (We have not found a way to compile the Analysis Component with MonoDevelop 4.0/Xamarin, which does not seem to support the .NET SDK 4.5.)
After installing the Mono Framework (see the instructions for Fedora), you should in theory be able to compile Visual Basic code in a terminal using vbnc, Mono's Visual Basic compiler, or with MonoDevelop. However, this does not work yet.
You need a collection of users preference sets in .ini format. See training data repository for details. This description assumes that you used the official training data repository. There are three important command line parameters you have to set for the analysis tool.
-pshould point to the folder storing the preferences. It will recursively take all .ini files as preferences.-cshould point to theconfig.inifile.-oshould point to the file (including its path) that represents the output of the analysis.
For example:
StatisticalAnalysis.exe -p "C:\GPII-Statistical-Matchmaker-Data" -c "C:\GPII-Statistical-Matchmaker-Data\config.ini" -o "C:\GPII-Statistical-Matchmaker\lib\StatisticalMatchMakerData.js"
The code of the GPII/Cloud4all Statistical Matchmaker's analysis component is available under the BSD 3-Clause License.
The research leading to these results has received funding from the European Union's Seventh Framework Programme (FP7/2007-2013) under grant agreement No.289016 (Cloud4all).