改进的全局K 均值算法及其在啤酒系统中的应用

K均值算法存在的问题一直限制其发展,主要问题在于:簇个数的确定、初始聚类中心选择和避免孤立点的问题。针对这些问题进行了改进优化,并把改进后的算法和动态递归模糊神经网络结合一起应用到了啤酒发酵系统当中。神经网络结构复杂,而粒子群算法可以优化全连接网络结构下的各层之间的连接权值和优化网络的拓扑结构。改进的粒子群优化算法也很大程度解决了早熟收敛的问题,有很好的泛化能力,在实际应用中改进的粒子群优化算法原理更简单,参数更少,实现更容易。

Improved Global K-Means and Its Application in Beer System

K-means algorithm has been limited by the main questions which are the problems to determine the number of clusters, initial cluster center points of selection and to avoid isolating the problem. To solve these problems the algorithm has been improved in this paper and the paper has applied the improved algorithm and dynamic recurrent fuzzy neural network to the beer fermentation systems. Because of complex neural network structure, the particle swarm optimization algorithm can be used to optimize connected network structure of the connection weights between layers and the network topology. This PSO does not easily trapped local minima and has better generalization ability. At the same time, in practical application the principle of improved PSO algorithm is simple and has less parameter so that it＇s easier to realize.