CI-based Clustering Technology And Its Application Research

Intelligence is some integrated capacities including purposeful acts of individuals, logical thinking and the ability to adapt to the environment. Computational Intelligence (CI) is a set of multiple intelligent methods, which specializes in reasoning and learning in the vague and inaccuracy environments. Thus, CI is a powerful computational tool for building intelligent systems and intelligent models.Clustering is an effective method especially for analyzing data so as to find useful information. Based on the simple idea:"Birds of a Feather Flock Together", clustering method divides data into several clusters by the rule that objects in same cluster have high similarities and objects between different clusters have not. Through the method, people can identify sparsely or intensive clustering regions, recognize global distribution patterns of data and interesting interrelationships between the attributes of data.CI is applied to clustering analysis according to establishing clustering analysis models and computation. The CI-based Clustering analysis models inherit both handling mechanisms and features of the biological systems. Therefore, it has the ability to naturally describe models in stead of establishing a precise mathematical model; it has a good ability to adapt to characteristics of objects; it has good self-organizing characteristics; it gains a good visualization of processing results, which facilitates understanding the results; it gains a certain intelligent capacity due to the intelligent features of CI; the variability and diversity of biological systems result in those of processing objects; it also an open system in that biological systems are easy to observe and analyze.CI consists of lots of fields. They are: fuzzy control, neural networks, evolutionary computation, swarm intelligence (SI), immunization algorithm, artificial life and DNA computation. In the paper, SOM and SI are chosen as two means for clustering. Based on them, we present dynamic SOM (DSOM) clustering model and self-organizing mixed ant colony (SOMAC) clustering model, and do some research on parallelizing in data parallelism and message-passing parallelism. Furthermore, DSOM and its parallel strategy are applied to customer classification, and SOMAC and its parallel strategy are applied to intrusion detection. Our experimental results demonstrate the effectiveness and the feasibility of above algorithms and strategies. The major research contents of the paper are summarized below:...