Date of Publication :7th October 2016
Abstract: In large number of real world applications the possibility of constituent volumes of unlabeled data is enormous. Moreover, the availability of labeled data is inadequate because of the expensive and annoying human interventions. The semisupervised learning is a substitute of supervised learning model utilizes the small amount of labeled data for training the massive volumes of unlabeled collections is an adequate model FOE enhancing the learners’ pursuance. In order to this Kai Zhang et al proposed a model that attempted to improve the Graph-Based Semi supervised Learning via Prototype Vector Machines. It uses scanty prototypes which are derived from data. Moreover, this mechanism will work effectively only on limited data samples. But, prediction of new data label from training data is more complex. The motivation gained from this model, an ensemble prototype vector machine for scaling classification performance that aimed to reduce the time and memory complexities of the kernel learning are used .The ensemble prototype vectors can handle large data sets without any complexity and for producing the new samples predictive analysis classification is performed on trained data. In predictive analysis, the decision trees are build on the training data for producing the new labels without any repeated factors. This ensemble model should achieve satisfactory classification performance.
Reference :