Create realistic social network from contact matrices in ABM

[elisabeth96]

The distribution of contacts of a person should follow the data that is available. Often, surveys give contact matrices between age groups in different contexts. A coefficient a_ij of a contact matrix gives the average number of contacts per day that a person from age group i has with persons from age group j. In SECIR, we use matrices for School, Work, Home, and Other (e.g. leisure).

For a given contact matrix, it should be possible to create realistic social networks, i.e. locations where the the age of the persons to which the location is assigned follows the same distribution as the contact matrix. The effective contacts of the location and probability/frequency of going to the location, e.g. in case of social events) needs to be tuned so the average total number of contacts per person per day matches the matrix. This will require experimentation.

We need an algorithm that, given a set of locations and persons and a contact matrix, assigns locations to persons and sets effective contacts. The number of persons per location should be variable to allow more or less mixing of social networks.

There are additional factors that could be taken into account, but since these effects are difficult to quantify, a basically random distribution would be fine for now:

• households: persons in the same household probably have a similar social network
• geography and social structure: persons that share one location are probably more likely to share other locations as well

[xsaschako]

@mknaranja, we think this could be a topic for a thesis, although not really suited for an issue anymore. Can this be closed?