[Link to the map and my blog post on it, which I’m reposting here](https://zarasophos.net/erasmus/). I recommend a desktop/laptop on fullscreen for the map and yes, it is sadly a bit laggy.
**What are you talking about?**
I created a map of a lot of the Erasmus student exchanges between 2008 and 2019 (a lot, not all, due to technical reasons explained below). The map is city-based, so all universities in one city are grouped together. You can also select time periods and also go through the data year by year.
The map only shows exchanges between Erasmus member states, so for example not the Eastern Partnership countries or Erasmus Mundus exchanges. It only shows student exchanges, not work placements (which would be another interesting map, but would require some further work).
While the data provided by the Commission is very nice, it doesn’t show geographic location. This means I had to geocode each exchange in the dataset. For this I had to start on two different levels: The first few years showed home and host universities, the other years cities. So, for the first few years, I first had to convert the codes of universities to the cities the universities were located in. Then I only had the names of cities, and those I could then geocode.
But this is of course much easier said then done: The data was hand-filled, so I had to deal with a lot of different ways of writing the names of cities. I did my best, but overall I was only able to identify around 84% of them. That sounds alright until you remember that an exchange involves two cities, so only 70% (0,84² = 0,7) of exchanges could be completely geocoded.
As you might notice, the map is already quite laggy because there are so many different exchanges on there. In order to help out a bit with that, I removed all exchanges that only included a single student.
In total, there are 4,9 million exchanges in the data. Out of those, 4,7 million were usable (mostly excluding non-student exchanges), and out of those, 3,3 million could successfully be geocoded. I couldn’t find data on how many people actually went on exchange between 2008 and 2013, but for 2014 to 2019, that means around 55% of all exchanges are included in the map.
Well, you probably shouldn’t. But if you are so inclined, it might be interesting to check the relations of cities: What cities do students in Marseille like to go on Erasmus to? How far does the Erasmus network of Sofia stretch to? And you can also check developments over time: How popular has Budapest been as an Erasmus destination? Has this changed, and if yes, when and why?
But that would just be my suggestions.
#POEPEN IN HET BUITENLAND
not in switzerland :d
Absolutely nothing to do with the weather. Or the parties. Nothing to see here.
to make things faster, maybe you can re-index to thousands of records instead of millions.
if you define every exchange as from a city to another city, and count all exchanges for that combination, you would get a number for every city combination.
take the lowest number, divide all numbers by that number, then keep only that number of connections for the map generation. does this make sense?
#gekoloniseerd
Could you pre-calculate the destinations to render the big lines once instead once per student? That should improve speed.
Well.. I went to Switzerland, Chur from Germany. I can’t find me 🙁
But Erasmus has been suspended this time so maybe that’s not in the dataset..
Great work, paying tribute to the EU’s most significant contribution to the continent. And just in time for Europe Day tomorrow, well done!
The net of the wealthy future woke leftist elite of the global governance.
11 comments
[Link to the map and my blog post on it, which I’m reposting here](https://zarasophos.net/erasmus/). I recommend a desktop/laptop on fullscreen for the map and yes, it is sadly a bit laggy.
**What are you talking about?**
I created a map of a lot of the Erasmus student exchanges between 2008 and 2019 (a lot, not all, due to technical reasons explained below). The map is city-based, so all universities in one city are grouped together. You can also select time periods and also go through the data year by year.
The map only shows exchanges between Erasmus member states, so for example not the Eastern Partnership countries or Erasmus Mundus exchanges. It only shows student exchanges, not work placements (which would be another interesting map, but would require some further work).
**Where did you get the data from?**
The [Erasmus data](https://data.europa.eu/data/datasets?query=erasmus%20mobility%20statistics&locale=en&publisher=http%3A%2F%2Fpublications.europa.eu%2Fresource%2Fauthority%2Fcorporate-body%2FEAC&page=1&limit=10) is published by the European Commission. Further data for geocoding the various universities is also taken [from the Commission](https://erasmus-plus.ec.europa.eu/document/higher-education-institutions-holding-an-eche-2021-2027) and [from Geonames](https://data.opendatasoft.com/explore/dataset/geonames-all-cities-with-a-population-1000%40public/table/?disjunctive.cou_name_en&sort=name).
**What did you do to the data?**
While the data provided by the Commission is very nice, it doesn’t show geographic location. This means I had to geocode each exchange in the dataset. For this I had to start on two different levels: The first few years showed home and host universities, the other years cities. So, for the first few years, I first had to convert the codes of universities to the cities the universities were located in. Then I only had the names of cities, and those I could then geocode.
But this is of course much easier said then done: The data was hand-filled, so I had to deal with a lot of different ways of writing the names of cities. I did my best, but overall I was only able to identify around 84% of them. That sounds alright until you remember that an exchange involves two cities, so only 70% (0,84² = 0,7) of exchanges could be completely geocoded.
As you might notice, the map is already quite laggy because there are so many different exchanges on there. In order to help out a bit with that, I removed all exchanges that only included a single student.
In total, there are 4,9 million exchanges in the data. Out of those, 4,7 million were usable (mostly excluding non-student exchanges), and out of those, 3,3 million could successfully be geocoded. I couldn’t find data on how many people actually went on exchange between 2008 and 2013, but for 2014 to 2019, that means around 55% of all exchanges are included in the map.
You can check what I did with the data yourself in this project’s [repository](https://github.com/maximilianhenning/erasmus-data).
**Why should I care about this?**
Well, you probably shouldn’t. But if you are so inclined, it might be interesting to check the relations of cities: What cities do students in Marseille like to go on Erasmus to? How far does the Erasmus network of Sofia stretch to? And you can also check developments over time: How popular has Budapest been as an Erasmus destination? Has this changed, and if yes, when and why?
But that would just be my suggestions.
#POEPEN IN HET BUITENLAND
not in switzerland :d
Absolutely nothing to do with the weather. Or the parties. Nothing to see here.
to make things faster, maybe you can re-index to thousands of records instead of millions.
if you define every exchange as from a city to another city, and count all exchanges for that combination, you would get a number for every city combination.
take the lowest number, divide all numbers by that number, then keep only that number of connections for the map generation. does this make sense?
#gekoloniseerd
Could you pre-calculate the destinations to render the big lines once instead once per student? That should improve speed.
Well.. I went to Switzerland, Chur from Germany. I can’t find me 🙁
But Erasmus has been suspended this time so maybe that’s not in the dataset..
Great work, paying tribute to the EU’s most significant contribution to the continent. And just in time for Europe Day tomorrow, well done!
The net of the wealthy future woke leftist elite of the global governance.
I went to Malta and it was great 2 weeks xd