uberwach/bundestag-votes

Data analysis on the voting patterns at the Bundestag (= German parliament).

Bundestag votes

Requirements

The build scripts are using many command line tools.

• curl and wget must be available from the command line.

• make and bash for the building.

• python 2 with typical data science libraries, if you want to redo the analysis.

• cvskit in particular in2csv is used for Excel to csv conversion.

• xmllint

Using make

make clean data will fetch the voting patterns as .xls files from the bundestag.de page, convert them to the .cvs format, and finally preprocesses/aggregates it into data/preprocessed/voting.csv which can be used for analysis.

Notebooks

Once voting.csv is present you can take a look at the notebooks, notebooks/Analysis.ipynb should be the most interesting so far.

Here is the visualization of the voting patterns, each dot is a member of the parliament colored in its factions. There is a small transparency, so overlaps are shown, too. From the $12$ votes so far a vector is build to have for each vote either the value: $1$ for "yes", $-1$ for "no", and "0" for abstained (being there and not passing the vote), corrupt, or absent (not being at the parliament at all). These $12$ dimensional vectors are then projected onto two dimensions via the t-SNE algorithm that gives more interesting insights than PCA in this case (in PCA you have the major axis being conformant-contrarian, and the next one - far weaker - is left-right).