INTEGRATION OF CLASSIFICATION TECHNIQUES UNDER SIZE CONSTRAINT: A CASE STUDY OF ELECTORAL DISTRICTING
Nattapong Musikauppatum, Surapong Uttama · 2012
The electoral districting is a problem of classification under size constraint. Its objective is to classify voters according to their traveling distances to fixed election units which are limited in capacity to support voters. Thus the goal of this paper is to propose a new classification algorithm to satisfy two conditions: size constraint and minimum traveling distance to election units. We modify k-nearest neighbor (k-NN) classification to support size constraint and combine with discriminant analysis. The experiment was tested on electoral districting of a district in Chiang Rai, Thailand. The result showed that our proposed algorithm satisfied the size constraint and the average traveling distance was 10% better than the manual classification and 0.82% better than modified k-NN and SVM with size constraint.