A0674
Title: Mixtures of local logistic regressions for nonlinear classification when data are heterogeneous
Authors: Hien Nguyen - University of Queensland (Australia) [presenting]
Abstract: Logistic regression has long been a staple method for the conduct of discrimination. Unfortunately, the standard logistic regression model only permits linear decision boundaries with respect to the input features of the model. We utilise the cluster-weighted modelling approach to construct logistic regression models that permit the expression of local rules and thus allow for nonlinear classification boundaries. These models are particularly suitable for the analysis of heterogeneous data, due the mixture construction.
