| As the application of wireless sensors'network becomes widely, clustering algorithms have been applied in this filed. Clustering algorithm is an important branch of machine learning and data mining. Once the clustering algorithm used in this filed, it can detect the changes of temperature, the movement of contamination, the leak of gas, and so on.The first algorithm DSE is based on data-centric distributed clustering algorithm. DSE algorithm is obtained through the Elink algorithm which has been improved. The main contribution of Elink algorithm is to select and use cluster head in orders. And this order embodies in the structure of quad-tree. The order of selecting cluster head cost the algorithm's time. The DES algorithm cancels this order, and introduces the virtual cluster head and the centre. When the cluster heads do not meet some distant metric, the algorithm can operate the heads synchronous. The time complexity of the improved algorithm is changed from O ( N×LogN) to O ( N).The second algorithm DHC is used to deal with time series which is based on DCDTW. In ICDM2008, DSIC algorithm also handles the time series. DSIC uses the K-Haar wavelet to compress the time sequence, and then uses the FASTDWT to compute the differences of time sequences. But this algorithm can induce big errors. DHC combines the SN-Haar wavelet and DTW, and adds the error compensate DC S.In this paper, DCDTW algorithm has been proposed. This method is used to deal with time series. At last, an anomaly detection method of IEEE 2006 has been applied to this algorithm.In this paper, two clustering algorithms which used in wireless sensor networks have been proposed. These two algorithms use different methods and frames for processing time series. DSE algorithm has the advantage of increased time complexity, but at the expense of clustering quality. DHC algorithm has the advantage of higher clustering quality and communication, but time complexity is its shortcoming. |
PDF Full Text Request