Abstract
Today there is massive applications that generate high volume of stream data like telecommunication systems and others. In this context, it is required to convert these data to valuable knowledge. Storing data stream to local storage and mining them can be considered as resource consuming process. Mining data streams means extracting valuable information and knowledge from continues data. This paper develops Adaptive sliding window random decision Tree (ASWRT). This model can learn adaptively from the changing data especially real time data that's related to the wireless sensors networks. The results of this research based mainly on the idea of data segmentation. This technique is one of the primary tasks of time series mining. This task is often used to generate interesting subsequence from a large time series sequence. Segmentation is one of the essential components in extracting significant patterns from time series data which may be useful in identifying the trend and changes in the prediction. The segmentations at ASWRT is mainly depending on mean and variance. The random decision tree has been employed as incremental builder for the tree for the purposes of classification. Other components to improve accuracy has been employed like sliding window-based algorithm and concept drafting detectors. ASWRT has achieved high accuracy and time performance over huge volume of data stream. These data generated by built-in random generator in MOA package. It achieved accuracy of 98.75% at time of 17.20 second in general.