The Theory Of Soft Clustering Over Date Stream And The Application Of Alarm In Gas Disaster

Aiming at the insufficient of existing clustering algorithm over data stream, a soft clustering algorithm over data stream based on cloud model is proposed and research is made from cloud synopsis data structure, soft clustering algorithm over data stream, the evolving analysis of soft clustering over data stream, the mining of the outlier, efficient evaluation of cluster and clustering over multiple data streams. The algorithm is applied to the gas disaster alarm and evaluation.A methods of abstract representation for data stream is proposed.The data with the weight vector of time decay is soft dividing by the cloud synopsis data structure and the cloud synopsis data structure is expressed, established and maintained.An algorithm of soft clustering over data stream based on cloud model is proposed. The extraction of data flow information summary is completed on the online data processing layer. The subordination of the data to the "soft mic_cluster" based on the degree the subordination is expressed;The final soft clustering results is found through the merging of the two adjacent with the max similarity on the offline data clustering layer. The application and performance test are made on real and artificial data sets. The experimental results show that the algorithm has good effect and high efficiencyExpanding research about soft clustering over data stream based on cloud model is made. The conception about the life cycle of cluster is presented and the evolving analysis from characteristics and increment is made;The concept of local similar outlier coefficient is put forward.The outlier is rapid detected through the local similar outlier coefficient and the causes of these outlier is analyzed; A clustering validity function is presented to overcome the limitations of existing evaluation indicators and achieves efficient evaluation for cluster.The clustering of the multiple data stream based on correlation is realized. The data cloud spot constructure with layer feature is introduced to expressed the different data sub_sequences. The clustering over the multiple data stream based on synchronization and asynchronization portion correlation is realized through the multiple layers cloud synopsis data structure. The experimental show that the algorithm has high quality and stability.The algorithm of soft clustering of data stream based on cloud model is applied to the gas disaster alarm and evaluation.The framework of the gas monitoring data stream clustering analysis application system is put; Pretreatment of the monitoring data stream of the gas is realized; The disaster alarm and evaluation of security level are made according to the gas monitoring data stream.The practical applications of the algorithm provids a processing method with application values for the questions about clustering with the feature of data stream.Lastly,the conclusion is made and the problems for further study are reviewed.