Research And Applications Of DNA Computing In Hierarchical Clustering Algorithm

As a new computing model, DNA computing uses DNA molecules to be the reactionmedium and takes advantage of biological techniques to finish the mathematical operation. Withthe problem domain mapped into particular DNA sequences, this novel paradigm obtains theoptimal combination, which satisfies all screening conditions, from the initial data pool through aseries of biochemical reactions. Keeping pace with the rising level of social informatization,more and more people need to find valuable information and knowledge from huge amounts ofdata. It makes data mining, the young and vibrant research field receive great attention byexperts at home and abroad. Acting as an important branch of data mining, clustering analysis isthe process of dividing a physical or abstract sample set into similar categories, and hierarchicalclustering algorithms aim to conduct stratification on objects and generate a spanning treeaccordingly. In recent years, they have been widely applied in many areas such as computerengineering, image processing and modern biology.Considering the high parallelism and low consumption of DNA computing, we introduce itinto hierarchical clustering algorithms and it’s possible to expand the scale of data sets andincrease the computing speed without changing the clustering quality. There are two main pointscontained this work, and the first one should be the SHCDM algorithm using sticker and2-armed DNA model. Note that in the2D plane, the single-linkage hierarchical clusteringalgorithm, which is based on the minimum-distance measurement, needs to be transformed intothe optimization problem of finding a minimum spanning tree in a weighted complete graph.Moreover, the outputs are affected by a given threshold when the iteration is terminated. It’s alsoimportant to point out that this algorithm maks full use of double-stranded structures, and theprocess of recognizing the best solution has strong operability and flexibility with the aid ofnanoparticle tagging DNA probes. As for some traditional clustering methods, we merge twogroups together in light of the closeness between a pair of points but neglect the neighborhoodinformation received from a global perspective. Faced with this situation, a fresh property called “link” should be proposed and applied to the similarity measurement of objects. According to theHCLDM algorithm based on Adleman and triple-stranded DNA model, two nodes associatedwith more links and higher correlation are clustered into one group preferentially, which presentsthe central idea of hierarchical clustering using categorical attributes and indicates the secondkey point in this paper. More precisely, nucleoprotein filaments play a pivotal role in forming thetriple-stranded structure and avoiding the mismatch effectively. The Adleman model constructsthe feasible solution space in one step, improving the reaction efficiency greatly and reducing theprobability of getting wrong results significantly.With the rapid development of Internet, a new trading pattern viewed as E-commerce hasmade a huge impact on community economy and brought about tremendous challenges andopportunities to traditional enterprises in Shandong province. Given the imbalance showed bySMEs (Small and Medium-sized Enterprises) during the development of E-commerce, theSHCDM algorithm could be used to cluster those enterprises with similar growth level and putforward suggestions combined with the questionnaire survey. In addition, group members havethe opportunity to build the strategic alliance, trying to design the development plan and profitmodel suitable for this particular type of businesses. In most social networks, the way offriendship building needs further innovations and the HCLDM algorithm achieves the division ofinviduals according to their closeness. After the classification, users’ resources are enriched andthe communication flexibility gets improvement successfully. As indicated in simulationexperiments, two DNA algorithms are both feasible and practical. Besides that, final clusteringresults have provided good solutions for the above problems, which are selected in accordancewith the evaluation function and actual conditions.