qedan / canadaelection2015 Goto Github PK

Implement a recommendation system for swaps for a voter from a specific riding and party. Should it maximize the total benefit of the swap? The benefit to that voter? Should it find a balance in benefit between the voter so that both are likely to agree to the transaction? Perhaps there is a way for a voter to indicate that they wont vote for a specific party and the system should only consider other recommendations.

I am currently assuming that the entire census population will vote in each riding. This is unrealistic because not all of those people are eligible voters, and not all eligible voters vote. What's a better way to estimate?

Create static visualizations that help make the ideas and data easy to understand.

It seems that the forecasts can have quite a bit of interesting time evolution, even over relatively short periods. This could be explored by saving snapshots of the forecasts, or getting historical data directly from threehundredeight.com somehow. Vote swapping is a futures market where the swappers agree to a trade on election day based on information they have today. So, looking at how the values of their swaps evolve in time is probably important.

After updating current_projections.txt there was a large change in the Lorenz curve. The gini coefficient went from 0.68 to 0.98. It is surprising that a few days difference in poll aggregations would cause such an extreme shift. Is this because of a bug somewhere? Is this consistent with the uncertainties in the system that are not yet being taken into account?

Create an interactive product that is presentable on the web.

Create interactive visualizations that allow a voter to set some parameters (checkboxes, sliders, map zoom, etc.) and see plots that are specifically relevant to them.

How often do you update current_projections.txt?

Could you have a quick description of those two terms or adding some references for further explanation?

The threehundredeight.com data has error values for their riding forecasts. These values will translate into errors on the value of each vote. Think about the math and implement the necessary functions.