Prototype Selection and Dimensionality Reduction on Multi-Label Data

Multi-label classification problem is one of the most general and relevant problems in the area of classification, where each item of the evaluated dataset is associated with more than one label. This paper discusses novel algorithms for prototype selection and dimensionality reduction on multi-label data. We have extended CNN (Condensed Nearest Neighbor) algorithm for multi-label data. We have also worked on an extension of the Class Augmented PCA(CA-PCA) method for multi-label data. These methods have been implemented on benchmark multi-label datasets and found to give good results.