A Review on Streaming Data and Decision Tree Classifiers for Non-stationary Data

Abstract

In streaming data, most of time the past data, does not make any sense for the prediction of current and future analysis of an event due to many circumstances and noise. This result into a poor prediction and less accurate model which happens because, the classifier is trained to work on the past data. In this paper, we are presenting the basic skeleton and existing methods in non-stationary type of data. This paper also discuss the weakness of current decision tree classifier i.e. problem associated with the searching of split point, memory size and running time complexity associated with classifier for non-stationary data as learning with small dataset and huge dataset is completely different. This paper gives an emphasis in the domain of speculations, statistical reasoning and forecasting to explore and understand the problems associated with the prediction model when it applies to a non-stationary data.