Clustering Data Streams over Sliding Windows by DCA

Mining data stream is a challenging research area in data mining, and concerns many applications. In stream models, the data is massive and evolving continuously, it can be read only once or a small number of times. Due to the limited memory availability, it is impossible to load the entire data set into memory. Traditional data mining techniques are not suitable for this kind of model and applications, and it is required to develop new approaches meeting these new paradigms. In this paper, we are interested in clustering data stream over sliding window. We investigate an efficient clustering algorithm based on DCA (Difference of Convex functions Algorithm). Comparative experiments with clustering using the standard K-means algorithm on some real-data sets are presented.