Research On Data Clustering Algorithm In Wireless Sensor Networks

With the development and application of networking technology as its core wireless sensor networks(Wireless Sensor Networks, referred to as WSNs) for its hardware and application flexibility and diversity have been widely used in various fields. However, the energy of the sensor nodes constitute WSNs limited and difficult to supplement restricts the popularization of the Internet of Things. So, how to from WSNs perceived huge amounts of data on the basis of the similarity of the data clustering in order to reduce a large number of redundant data transmission, reduce the energy consumption of sensor nodes and prolong the network life is a practical problem facing WSNs applications.Therefore, the study of this topic is undoubtedly of great practical significance and application value, and was funded by the National Science and Technology Pillar Program during the Twelfth Five-year Plan Period(No.2014BAF07B01).On the basis of consulting a large amount of chinese and foreign literature, the combination of WSNs sensor nodes and sensing data with random distribution and the characteristics of dual attribute, wireless sensor network data clustering algorithm was studied.The specific research work are:(1) Research and analysis of the existing classical clustering algorithm and conducted a performance analysis and comparison of these algorithms;.(2) Through the study of wireless sensor network-related knowledge, in-depth analysis of the similarities and differences of the wireless sensor network data and image data, research and analysis of the clustering method for wireless sensing data, and according to the data characteristics of wireless sensor networks, introduced the data field theory, research and analysis of the clustering properties of the data field;(3) Proposed a data-field based wireless sensor networks dual clustering algorithm. First, by analyzing the information entropy theory, combined with the information entropy for sensor nodes data given the appropriate weights; secondly reference data field theory describe the interaction between objects, considering the physical space object attributes and information space attributes, to achieve the nonlinear mapping from the feature space scene to potential space of sensor nodes data; finally according to the field theory in the same potential value of potential line or equipotential surface form a natural cluster characteristics, the data objects to be divided. And the partitioning phase is completed mainly based on the potential center to find and data partition through routh and detailed;(4) Experimental simulation by synthetic data and UCI real data sets, compared with the experimental results and performance analysis, shows that the dual algorithm for clustering data of wireless sensor network based on data field has better clustering performance and good scalability and low computational complexity.