The Research And Application Of Fuzzy Clustering Based On Improved Genetic Algorithm

Objective function based clustering algorithm, fuzzy C-means clustering algorithm theory the most perfect, the most widely used. In theory, it's hiking through iterative technique to find the optimal solution is a local search algorithm. Therefore it contains an obvious drawback, that is, vulnerable to the impact of the initial value to be trapped in local minima. Genetic algorithm is a widely used global optimization algorithm, which is simple, universal, robust to noise and other characteristics, is a question not related with the algorithm for solving model. Because of these advantages of genetic algorithm can solve the fuzzy C-means clustering algorithm to initialize the sensitive issue. Therefore, the fuzzy C-means clustering algorithm and genetic algorithm together with the use of both can play Fuzzy C-means clustering algorithm the local search ability and full consideration of the genetic algorithm global search capability, thereby improving the convergence speed of hybrid and better address the clustering problem.By reading a lot of literature, and fuzzy clustering, genetic algorithms and understand the absorption of other related algorithms and research, this paper presents a genetic algorithm based on improved fuzzy C-means clustering algorithm. The major work as follows:Firstly, the improvement of basic genetic algorithm. In the genetic algorithm to each individual based on the best seeds from the current population divided into the dominant population, the second best two populations, and adopt different strategies on the genetic evolution of two species evolved separately. In the choice of strategy, the use of elitism and roulette mixed strategy, and different from the past is to allow the next generation of elite individuals involved in genetic manipulation, thus ensuring the convergence of the algorithm to ensure the stability of genetic evolution, inhibition of invalid solution the diffusion, the center of the cluster search efficiency. Crossover and mutation, the dominant species based mainly on cross-, second-best populations to variation mainly to ensure the population's average fitness and population diversity.Secondly, the improved genetic algorithm to solve initial value fuzzy C-means clustering on sensitive issues. This algorithm uses the genetic algorithm fuzzy C-means clustering algorithm to optimize the initial cluster centers to solve the initial value of the fuzzy clustering algorithm. To solve the problem, encoded using the cluster centers as the real-coded chromosome mechanism, this representation makes the search space to expand, is conducive to global search, and solving accuracy improved. Fitness function by means of Fuzzy C-means clustering algorithm the objective function. Criteria used to determine the maximum number of iterations and up and down several generations of changes in the average fitness value is less than a threshold to determine, reduce the running time of the genetic algorithm.Thirdly, the performance analysis of improved algorithm. In MATLAB 7.0 for the experimental simulation platform, using a standard data set IRIS comparison test fuzzy C-means clustering algorithm, simple genetic algorithm based on fuzzy C-means clustering algorithm and the proposed improved genetic algorithm based on fuzzy C-means clustering performance of the algorithm, experimental results show that the algorithm can overcome the fuzzy C-means clustering algorithm sensitive to the initialization of the shortcomings of the average number of iterations and accuracy has certain advantages.Fourthly, the application of. improved algorithm The algorithms used in the paper analysis system for processing examination results, the fuzzy clustering results reflect the significance and role of the Teaching and Learning.