In Switzerland, La Poste has a legal obligation to make sure that everybody has access to a close-by post office. In some remote areas though, La Poste has invented a system to provide postal services directly at home through the mailman, who will ring the bell at your door if you have place a little sign on your mailbox.
The number of villages where this home-delivery service is in place is now a third of all postal access points in Switzerland. While La Poste is saving money by shutting down post offices and replacing them with home-delivery systems, people who aren’t staying at home all day long (like unemployed or retired ones) can’t easily send a package or pay a bill.
La Poste refused to give me the complete list of these villages so I decided to scrape all the pointon the map that they provide (which is not the best for visualization).
As of August 26th, 2015, it looks like La Poste is migrating its websites to a new platform, and the map that we are interested in is still up, but doesn’t seem to appear on the new website (see notebook for links).
In the last weeks, it also seemed like the map was undergoing lost of maintenance work: the query url parameters have been modified; the number of villages has increased by two; one error point with coordinates in Somalia was removed; and layers of security certificates have been added.
These changes have slowed me down significantly as I believed that the code was creating mistakes, but I added code for every possible mistake that could turn up.
After scraping the 1245 points (json objects) and their coordinates on the map and examining them, we noticed that 41 of them were structured differently and had to clean the data to produce a dataframe that we could work on.
We tried to add columns to our dataframe with population data, but we realized that the villages list was too granualar (it included ‘hamlets’ that didn’t even have zip codes, for example) to be matched with any other database.
By filling the missing cells by hand, we were able to add column containing the canton in which the village was located (in cartodb with an SQL query), to finally produce a clean map with all concerned villages and thir canton.
This will help us demonstrate what region are more concerned by the problem. The next step will be to join this dataset with a population by hectares dataset, to see if the system creates inequalities, which is our hypothesis.