Functional logistic regression with fused lasso penalty

Abstract

This thesis considers the binary classification of functional data collected in the form of curves. In particular, we assume the situation when the functional predictors are highly mixed over the entire domain, so that global discriminant analysis that is based on the entire domain is not effective. To address this problem, this thesis proposes an interval-based classification method for functional data

the informative intervals for classification are selected and used for separating the curves into two classes. The proposed method, called functional logistic regression with fused lasso penalty (FLR-FLP), combines the functional logistic regression as a classifier and the fused lasso penalty for selecting discriminant segments. FLR-FLP automatically selects the most informative segments of functional data for classification via the fused lasso penalty, and simultaneously classifies the data based on the selected segments via the functional logistic regression. The effectiveness of the proposed method is demonstrated with simulated and real data examples.