Mathematical Programming for Automatic Dwelling Segmentation : The 2010 Argentinian Census in Buenos Aires Province as a Case of Study

Abstract

Planning a population census involves several logistical challenges, one of which is to determine which homes should each enumerator visit. This is known as the dwelling segmentation problem, which usually involves certain constraints on the assigned route of each enumerator. During the 2010 National Census in Argentina, a solution for this problem based on integer linear programming was successfully applied in the Province of Buenos Aires. This work discusses the proposed solution, identifies several problems and develops alternative mechanisms for correcting them, getting even more satisfactory solutions. Moreover, new approaches for the problem are suggested and the economic impact is analyzed, getting a potential reduction of 6.96% of the total payment of the enumerators, compared to the original solution. These improvements, due to the way of finding feasible solutions, have not worsened the quality of the solutions found in comparison with the original problem, making it worthwhile to be implemented.