B1342
Title: Curve classification based on feature weighted models with application to medical data
Authors: Chunzheng Cao - Nanjing University of Information Science and Technology (China) [presenting]
Abstract: The classification of complex functional data with small inter-class differences and large intra-class differences in the local features is studied. A new type of feature weighted method is proposed. By constructing appropriate weighting functions, it is possible to amplify important features and reduce the interference of bad features on classification. The contribution regards some feature weighted classification by means of F statistics and the weight function based on mean-variance models to improve the accuracy of classification for complex medical data.
