Embedding preference orders in 2D Euclidean space

75 Views Asked by At

PROBLEM

I have a set of candidates $C=\{c_1,c_2,...,c_m\}$ and a set of voters $V=\{v_1,v_2,...,v_n\}$. Each voter ranks all of the candidates (e.g. $v_1: c_1 \succ c_2 \succ c_3 \succ \dots \succ c_m$).

Now I would like to embed both voters and candidates in a 2D Euclidean space. Each voter and each candidate should be represented as a single point (therefore, I will have $m+n$ points).

MY IDEA

First, we create a neighborhood matrix based on voters' preference orders. Then, we use some algorithm (e.g. PCA) to reduce the dimensions. Finally, we add candidates' points as weighted average of all voters.

QUESTION

Although my method is giving satisfying results, I am very curious about some known methods. I searched in Google and Google Scholar but I cannot find anything interesting enough. Can you recommend me where I can search for known solutions to this problem?