Closed mgarod closed 8 years ago
We could do two things: In the last paper on this, I started with the highest-level centers (Oberzentren) as input, and then tried to find the lower-level centers (Mittelzentren) around them. But ultimately, it would be great to actually identify them automatically. I'm thinking a combination of number of incoming links (not just geolinks!) plus information about the type of entity (to make sure we don't classify a museum as a center just because it has a lot of incoming links and a georeference). Does that make sense?
I think I understand. In the older method, we'd run with all highest-level centers, and taking the top 6 incoming for each, we should obtain the complete list of all lower-level centers from the official list. And then this is where precision and recall would be used for accuracy of misclassification. In this way we wouldn't have a way to classify the high-level centers, right?
Yep, exactly.
I'm not sure that I understand how we will be classifying München as Oberzentren, or Freidberg as Mittelzentren?
How will we be using the number of incoming geolinks?