Research Of Clonal Network Clustering Algorithm Based On Computational Intelligence And Its Application

Data Mining Technology, as a new emerging advantageous method, can automatically analyze magnanimity datum resources and transform it to the significant knowledge. Clustering Analysis as a key technology in the domain of Data Mining, gradually has become the hot and challenging research topic in recent years. With using mathematic methods in researching and processing of object datum, Clustering Analysis can classify the given objects by comparing the similarities between them without teacher's guidance in this course, so it belongs to the unsupervised classification.This artical has carried on deeper research to existing clustering algorithms. First of all, based on analysis the existing traditional Clustering Algorithms, in view of the traditional algorithm insufficiency, also closely tiding with the viewpoint of intelligent fusion and supplementary, this article focus on one unique clustering algorithm which based on intelligent fusion technology. Intelligent computation has the good optimized characteristic and self-study,auto-adapted ability. With introducing intelligent computation method into clustering analysis, the new achieving intelligent algorithm doesn't only can overcome the insufficiency of traditional ones, but also has higher cluster's validity and usability.Followed, based on gathering,research and analysis the existing traditional Clustering Algorithm, in view of the traditional algorithm insufficiency, this article has proposed a new kind of clustering algorithm which based on intelligent fusion technology. This algorithm uses the immunity clonal strategy into the network architecture, combining with mutation operator and the forbidden clonal operation together to train the given original data for study, further evolves a clonal network to reflect the original data in the state space. This kind can not merely overcome the general method to the shortcoming of initializing sensitivly , relying on cluster's prototype or disappearing slowly, but effectively while distributed having nothing to do with the given data , also can deal with the network data with complicated magnanimity , multidimension or multiattribution . This article passes the validity of the verification algorithm of emulation with strict experiment . With using this unsupervised clustering algorithm for abnormal intrusion detection, a new abnormal intrusion detective algorithm is proposed based on clonal network clustering. The computer simulation on the KDD CUP99 dataset shows that this method is feasible and effective.