Classification of Shape Data Using Logistic Regression

In this paper, a linear model using the principal component scores in shape data that is fitted in the tangent space of shapes with the nominal responses is proposed. In that respect, the multinomial logistic regression for multivariate and logistic regression for binary responses is considered. In doing so, Principal Components in the tangent space are used to improve the estimation of logistic model parameters under multi-collinearity and reduce the input data dimension. The result of this study is significant and point to improving the classification shape data according to their different nominal groups. We study the performance of the proposed technique on three real-world data sets.