Research And Application Of DNA Computing In Cluster Analysis

Data mining applies kinds of methods in order to extract data patterns on user interest. It is the essentialstep of Knowledge Discovery in Databases and one of the most important functions of it is cluster analysis.Clustering is the process of grouping a set of physical or abstract objects into classes of similar objects. It is theeffective means for people to understand and explore the intrinsic link between things. At present, study onclustering analysis is active and many clustering algorithms are put forward. However, these algorithms arebased on silicon computer. With the rapid development of electronics computer technology, the computer’scomputing ability comes to its physical limit. Because of being restricted by the computing capability, clusteranalysis shows its powerless in the process of large-scale data. DNA computation is a new computing methodand it can do much computing and parallel computing with fast speed. So if DNA computation is applied tocluster analysis, it will satisfy the amount of clustering’s computing, and it will accelerate the speed of thecluster, increasing the size of the data, and reduce the complexity of clustering.The basic principle of DNA computing is the double helix structure of DNA molecules and the WC basepairing coding principles, and then the practical issues will be mapped to the molecular chain of DNA, withthe feasibility of the role of enzyme biological operations to generate all the data pool. And then followingabstract issue specific rules the mapping of DNA molecules will be operated by biochemical controlledreactions. Finally, using molecular biological techniques (such as affinity chromatography, polymerase chainreaction PCR cloning, bead separation, mutagenesis, electrophoresis, molecular purification, etc.), to detect thepossible result of the operation.This paper presents the method of applying DNA computing into cluster analysis, the main idea aredivided into two directions: one is that directly use the super computing power of DNA computing in clusteranalysis, and the clustering problem is mapped to the biological operational problems of the DNA molecules;the second is to optimize existing cluster analysis method to reduce the amount of the calculation of thesemethods, and improve the clustering speed. Firstly, we convert clustering problems into DNA computingproblems by clever design of DNA encoding, and we complete the clustering process in a test tube through thesticker model. Secondly, we introduce DNA computing to existing spectral clustering, under the premise that not losing the advantages of the existing clustering algorithms, and reduces complexity, increases theclustering speed. After the simulation test through mat lab with the giving data set, results show that these twoideas are valid in theory. However, because of being based on DNA computing, compared with the generalapproach, the method’s advantage of parallel speed is cannot be compared. Of course, under the exitingbiological level, DNA computing is still relatively complex, but with the level of biological technologycontinues to mature, the advantages of DNA computing will become increasingly apparent. The inadequaciesof this article is: first, by a variety of conditions, it hasn’t made actual validation of the proposed algorithm inthe biological laboratory; second, DNA computing in the optimization of the existing spectral clusteringalgorithms, optimization is not enough, failed to fully demonstrate the advantages of DNA computing.This paper presents the application of DNA computing in the cluster analysis, and simulation results showthe feasibility of the methods. It provides some help and some foundation to the research of DNA computingusing in the field of data mining and clustering analysis.