Knowledge Transfer Using Cost Sensitive Online Learning Classification

A real assumption in numerous machine learning area and data mining algorithms is that the future data and training necessary to be in the similar feature space and have the similar distribution. Moreover, in numerous applications, this presumption may not hold. For illustration, we once in a while have task of a classification in one space of interest, yet we just have sufficient preparing data in an alternate area of interest, where the recent data may be in an alternate feature space or take after an alternate distribution of data. In such cases, knowledge transfer, if done effectively, would extraordinarily enhance the execution of learning by staying away from much costly data-naming efforts. Recently, transfer learning has developed as another learning structure to address this issue. Another machine learning structure called Online Transfer Learning (OTL) that plans to transfer information from some source space to an online learning assignment on a target domain. A famous methodology to cost-sensitive learning is to rescale the classes as per their mis-classification costs. In spite of the fact that this methodology is successful in managing with binary class issues, late studies demonstrate that it is frequently not all that supportive while being connected to multi-class issues specifically. In this paper, we have focus the survey on cost sensitive on machine learning and various methods used. This paper also focuses on online learning methods.