Data mining techniques for data streams mining

Abstract

In resent years data stream mining plays an important role in real-time applications that generate gigantic of data needed intelligent data processing and on-line data analysis . The source of high-speed data streams may include video surveillance systems, stock markets, internet traffic, tweets etc. Traditional data mining techniques can’t feasible for the data stream mining due to unique characteristics of data streams such as high dimensional, continuous flow, high-speed and fast changing. It necessitates building new data mining techniques or modifying existing ones to mine data streams. The main challenges include that the data stream mining needs to handle data distribution and concept drifting. This paper analyzes the challenges involved in designing data mining techniques for mining data streams besides evaluating various existing techniques and their preprocessing methods. The evaluation results reveal which methods are feasible and which methods are not feasible in real-time data streaming applications.