Pengkajian algoritma exchange untuk analisis regresi linear bergerombol dengan metode kuadrat terkecil

Abstract

Nonlinear relationships and heterogeneous subjects are common problems in regression. These problems may happen because of many causes, such as due to the fact that data consist of several groups which each group has specific regression function. Unfortunately, in most cases, it is not known a priori which subset of observations should be approximated with which specific regression function. Clusterwise linear regression by Ordinary least square approach which is implemented with exchange algorithm is one of the proven method that can be implemented to find the optimal clusters that can optimize the objective function. This method is recommended to overcome the nonlinearity and heterogeneity of existing data. The weakness of this method is that the more number of cluster and observation are, the longer likely to compute. And there is still miss clustering on a data set which has an overlapping observations.