Survey of Novel Process for Domain Adaptation and Cost Sensitive Online Classification in Data Mining

In the society of machine learning and data mining, online learning and cost-sensitive classification are two topics which are very widely studied. Though those topics are very widely studied but important problem like Cost-Sensitive Online Classification is studied very inadequately. In age of big data, there is an urgent need of developing a technique to mine fast increasing data. There are many methods which are extensively studied for machine learning. Main aim of online learning is learn more forecast model which will make correct forecasts/predictions on the flow of examples which will arrive consecutively. Online learning is beneficial because it has high efficiency and scalability. Online learning has been used for solving online classification tasks. These tasks include a wide range of real-world data mining applications. Many Online learning techniques have been implemented for online classification tasks. For example: Perceptron algorithm, Passive-aggressive (PA) learning. Although it is broadly studied, already implemented techniques are not appropriate for cost-sensitive classification tasks. Misclassification costs need to be focused in cost-sensitive classification tasks. Most of already implemented online learning methods are depended on online classification algorithm.