This is the abstract and slides of my talk about Political Dynamics of the Mexican Senate at the Conference of Complex Systems’ satellite Computational Social Sciences.
With the vast amounts of data available freely about virtually any field of knowledge, one of the greatest challenges for today’s scientists is to be able to store, organize and analyze this data and to use it for novel and useful applications.
Political sciences and legislation are fields that have seen such an increase. The Mexican government and several of its dependencies have made a lot of their databases publicly available online, with the Chamber of Senators being the main focus of this work.
New oportunities to be aware of the actions that decision makers are taking are arising and showing if a real representative democracy is being held. In this work we present a framework for automatic data acquisition, construction of a graph oriented data base and statistical modeling of data taking advantage of the capabilities of cloud computing.
The data collection was though the Mexican Senate’s official website, so everything is completely open. The information gathered consist of the names of the senators and their alternates, party and comissions they belong to, entity they represent, the edicts, how did the senators vote, attendance and the dates in which the above happened. Followed by this a graph oriented database was build which allows to performm an analysis of the senators actions and find communities in a temporal basis. A distance matrix between each senator was created from the votes which was used to perform statistical analysis such as multidimensional scaling for the projection of the vectors asociated with the senators and the construction of a weighted network in order to find communities amog them and study it’s topological properties. Another network was also built from the joint proposals of the edicts by the commissions because each edict is proposed by one or more commissions. A new analysis of communities was carried out for this network, finding 3 great subjects that after a manual review we determined that they reflect the following topics; governance, foreing affairs and social issues.
The final part of the paper is to determine if it is possible to predict the vote of a particular senator through his or her history and the metadata we have. Obtaining the best accuracy by means of logistic regression with a value of 0.7082, surpassing the 0.6608 of predicting that they always vote for pro.