International Journal of Advanced Research in Computer Science and Software Engineering (IJARCSSE)
The concept-drift phenomenon is used for outlier detection or data labeling, which plays a vital role in detection of outlier. But in that there is a disadvantage which is of re-clustering when drift occurred. In this connection two scanning operations are required, one for the drifting and another for the re-clustering of sliding window. It is necessary to investigate the principal of clustering to design efficient algorithms to minimize the disk I/O and minimizing the number of scanning operations. In this paper, to overcome the problems of scanning operations and also it is extended to the categorical data where as in literature the leader algorithm for the numerical domain and sequence of data set.