Clustering Analysis Based On Swarm Intelligence Optimization Algorithm

Clustering Analysis (CA) is a kind of unsupervised classification, which is divided into different classes of data or objects by a set of specific rules. Compared with szupervised classification, although there is no higher classification accuracy,but it does not require a priori knowledge, it is Polymerization around the internal data, in the practical application can get better application. However, due .to the initial cluster centers clustering of CA was sensitive, and it easy to fall into local optimum. To solve this problem, Many improved methods been proposed. Swarm Intelligence Optimization Algorithm Cluster Analysis(SIOACA) is one of the important research direction, which is the clustering problem as an optimization problem, then heuristic search through the global parallel search.In this paper, the research focused on SIOACA, through analyzed method which conventional CA and classical SIOACA, to study its process and existing problems,aiming sensitive issue for CA, two new types of SIOACA are proposed, that is CA based on Fireworks Algorithm (FACA) and Clustering Analysis based on Hybrid Encoding Mode(HEMCA). And through it compared with the traditional CA method and Classical SIOACA by evaluated performance.FACA was fireworks algorithm which a new type of swarm intelligent optimization algorithm (SIOA) applied to CA, the algorithm combines two kinds of different search strategies, respectively using Real Number Code (RNC) and Binary Code, Fireworks Clustering Algorithm (FCA) based on two kinds of encoding was proposed, and through the simulation experiments, we analyzed the performance of two algorithms for different ways,through the experimental results, the fireworks algorithm based on binary encoding has good clustering effect, high stability, and the classification accuracy is higher than Classical SIOACA.HEMCA is mixed with Encoding Mode based on Clustering Center (EMCC) and Encoding Mode based on Sample Number (EMSN), Quantum Particle Swarm Optimization(QPSO) and Improved Rain Forest Algorithm (IRFA) were used to CA by different encoding mode. Through simulation experiment, in the search process, EMCC is easy to produce solutions that exceed the search space, so as make the search into local optimum. When the search space is fixed,while EMSN was used,although it is easy to control the search range,but it limits the scope of the search space, which is not conducive to the further improvement the quality of the optimal solution. HEMCA not only solves the problem of beyond search space, but also can keep diversity of population, and through experimental comparison,Classification Accuracy is better than the traditional clustering analysis method.