美国数学建模竞赛特等奖论文——2007 A O Z What to Feed a Gerrymander【数学建模】.pdfVIP

232 The UMAPJournal 28.3 (2007) What to Feed a Gerrymander Ben Conlee Abe Othman Chris Yetter Harvard University Cambridge, MA Advisor: Clifford H. Taubes Summary Gerrymandering, the practice of dividing political districts into winding and unfair geometries, has a deleterious effect on democratic accountability and participation. Incumbent politicians have an incentive to create districts to their advantage (California in 2000, Texas in 2003); so one proposed remedy for gerrymandering is to adopt an objective, possibly computerized, methodology for districting. We present two efficient algorithms for solving the districting problem by modeling it as a Markov decision process that rewards traditional measures of district goodness: equality of population, continuity, preservation of county lines, and compactness of shape. Our Multi-Seeded Growth Model simulates the creation of a fixed number of districts for an arbitrary geography by plant- ing seeds for districts and specifying particular growth rules. The result of this process is refined in our Partition Optimization Model, which uses stochas- tic domain hill-climbing to make small changes in district lines to improve goodness. We include as an extension an optimization to minimize projected inequality in district populations between redistrictings. As a case study, we implement our models to create an unbiased, geo- graphically simple districting of New York using tract-level data from the 2000 Census. The text of this paper appears on pp. 261-280.