DMCS: Dual-Model Classification System and Its Application in Medicine

摘要

The paper introduces a novel dual-model classification method -Dual-Model Classification System (DMCS). The DMCS is a personalized or transductive system which is created for every new input vector and trained on a small number of data. These data are selected from the whole training data set and they are closest to the new vector in the input space. In the proposed DMCS, two transductive fuzzy inference models are taken as the structure functions and trained with different sub-training data sets. In this paper, DMCS is illustrated on a case study: a real medical decision support problem of estimating the survival of hemodialysis patients. This personalized modeling method can also be applied to solve other classification problems.