Study On Clustering Algorithm Based On The System Energy Theory And Its Application

Clustering is to assemble the data which have the similar character from the huge data set, and to class the data which have different character, and make the comparability in different class as small as possible, but as big as possible in one class. Today, clustering algorithm has been used in a lot of fields, but with the progress of the human's cognition of the nature and society, this has brought forward a challenge to the existed clustering algorithm. Without doubt, these algorithms can deal with some problems, but there isn't an almighty technique. It is a new study field to use clustering algorithm into intrusion detection system. But with the high speed development of internet, it has brought forward a lot of new requirements to intrusion detection system. With these challenge and requirement, it needs humanity to make further study.The paper firstly presents the definition of clustering algorithm and summaries several kinds of common clustering algorithm and analyses the validation of them, and lists several criterions of validation. Then, the paper brings forward a new clustering algorithm, which called Clustering Algorithm Based on the System Energy. During this process, the paper defines the energy function and pretreat process, and describes the clustering process in detail, and integrate the theory of entropy defines a new clustering validation criterion. Moreover, through the relationship between information entropy and distribution of swatch, information entropy and comparability coefficient of swatch, comparability coefficient of swatch and swatch system energy deduced the relationship between the theory of entropy and swatch system energy, and through the experiment prove the validation of clustering algorithm and clustering validation citerion. After these steps, for the sake of getting system energy more exactly and enhancing the veracity of clustering algorithm, the paper through analyses the neural network, especially the BP neural network, brings forward a new network model. Uses this network model to adjust the ratio of different attribution, and analyses the way of defining the initialized ratio and the process of swatch treatment. When it is finished, the paper summarizes the principle of intrusion detection system and its capability evaluates method, designs the system frame, and brings forward the detection process, and analyses the data reduction method to improve the detection efficiency. At last, the paper uses the experiment on the KDD99 data set to prove that the algorithm can be faultlessly used in intrusion detection system, and the improvement of neural network make the veracity of clustering and the efficiency of intrusion detection system greatly increased.The paper uses system energy theory into clustering algorithm and uses the algorithm into intrusion detection system can faultlessly clustering and detection. Especially with the improvement of neural network, making clustering and network intrusion detection more exact, this will be propitious to the further study of clustering algorithm and network intrusion detection.