Abstract:

In recent years, the use of neural network algorithms to solve clustering problems has been a hot topic. The Elastic Network Algorithm（ENA） is a powerful neural network algorithm, but it is mainly used for TSP problems and it rarely be used for clustering problems. In this paper, the structure of the elastic network is changed after analyzing the characteristics of the elastic network and clustering, then the Elastic Network Algorithm with Center Movement for clustering（CMENA） is proposed. The energy function of the elastic network is adjusted with the objective function of the cluster. By minimizing the energy function, it controls the movement of the cluster center to obtain the clustering results. It has the advantages of tracking the clustering process and stable clustering results. Through a large number of experiments, it has been proved that the clustering results of this algorithm are unified multiple times. Compared with other clustering algorithms, the clustering quality is significantly improved.

Key words: elastic network, cluster analysis, center move, incremental control

摘要：

运用神经网络算法求解聚类问题是近年来的研究热点。弹性网络算法（ENA）是一种强大的神经网络算法，但其主要用于旅行商问题，很少用于求解聚类问题。面向聚类问题的特点，调整并优化了弹性网络的结构，提出了具有中心移动特性的弹性网络聚类算法（CMENA）。该算法依据聚类的目标调整并优化了弹性网络的能量函数，通过新能量函数的最小化，控制聚类中心神经元的移动，得到聚类结果，具有聚类过程可跟踪，聚类结果稳定等优点。通过大量实验证明，该算法聚类结果统一，与其他常用聚类算法相比，聚类效果显著提高。

关键词: 弹性网络, 聚类分析, 中心移动, 增量控制