The Research And Realization Of Clustering Algorithm In Data Streams Mining

Cluster analysis is an important topic in data mining field. With the fast development of computer and application , people have more strong ability to obtain data streams. Being an important resource, data streams have been payed more attention,such as web click streams, meteorological observation information Streams, telephone record streams etc. Compared with the traditional data,these data are continuous, dynamic, variational and boundless. We can only visit them once or several limited times and only store the summary dynamicly. These natures make data streams mining become more difficulty, and it has also brought about higher request for data streams cluster-algorithm.In recent years,a lot of clustering algorithms for data streams have been proposed and applied. In this thesis, firstly ,we introduce the relevant algorithm and technology,and then state the nature of data streams mining.Finally, we present a new density-based clustering algorithm over an evolving data streams-Sdstream. The proposed algorithm uses sliding window technology and adopts pruning strategy.We can not only find clusterings with arbitrary shape and amount, but also deal with noise data,cut down memory cost, query and analyze history informations by using this algorithm. It is an effect data streams clustering algorithm.The experimental results on real datasets and synthetic datasets demonstrate that the algorithm has good applicability, effectiveness and scalability. The algorithm is suitable for dealing with large-scale evolving data streams .