Font Size: a A A

Study On Distributed Data Mining On Wireless Sensors Network

Posted on:2009-05-28Degree:MasterType:Thesis
Country:ChinaCandidate:L ChenFull Text:PDF
GTID:2178360242476655Subject:Pattern Recognition and Intelligent Systems
Abstract/Summary:PDF Full Text Request
Wireless Sensors Network ( WSN ) combines sensor technology, communication technology and computer technology together. It has the ability of collecting, transferring and processing data. With the quick growth of the wireless network technology, it has been used in almost all of fields. We need handle the huge amount of data in the WSN with data mining technology. In this paper,we focus on the distributed data mining's application in WSN based on the characteristics of WSN.First, this paper sums up the intrusion detection technology and distributed data mining technology, then introduce the current situation and major algorithms. Second, we provide two approaches in intrusion detection. At last, we introduce an energy efficiency distributed data query method for WSN.Based on the previous work of intrusion detection and distributed data mining, the major contributions of the paper are as follows:(1)To the mistake of local optimization in iterative process of fuzzy C-means(FCM),we present a clustering algorithm based on ant algorithm to identify fuzzy clustering numbers and initial clustering center of FCM . Then we apply the approach to detect the intrusion in WSN.(2)Traditional intrusion detection method focused on detecting intrusions in a continuous attribute space. The categorical attributes are typically ignored or incorrectly modeled by existing approaches. On the other hand, traditional methods can't analyze the dynamic data or stream data .Given consideration in the nature of mixed-attribute data we present a tunable algorithm for distributed outlier detection in mixed-attribute data sets to address these challenges.(3)For inquiry data in the WSN, a new energy efficiency data inquiry approach which is based on learning theory is proposed. In a cluster all nodes select head-node by energy principal, the head-node uses kernel function to fit other nodes in the cluster. By the approach, we can not only get the data from WSN at certain precision, but also extend the lifetime of the WSN.
Keywords/Search Tags:wireless sensors network, distributed data mining, intrusion detection, Ant Colony, mixed-attribute data, energy efficiency
PDF Full Text Request
Related items