Pre-clustering algorithm for anomaly detection and clustering that uses variable size buckets

Clustering is known as grouping of data based on their similarities. This paper introduces an algorithm of k means for clustering of data streams and detection of outliers. The introduced technique for detection of outliers is based on distance as well as on time on which they arrive in the cluster. This paper also takes into account the selection of k centers and variable size of buckets with the help of which space can be effectively utilized during clustering. Most traditional algorithms make clustering a very difficult problem by reducing their quality for a better efficiency. This paper indicates that with a small increase in time you can efficiently cluster the data without much loss of quality of data.