具有加权特性的弹性网络聚类算法研究

聚类分析是数据挖掘中重要内容之一，也是人们分析数据的重要工具。针对聚类分析中存在易受噪声干扰、高维数据聚类结果不佳等问题，对弹性网络进行了加权聚类方向的研究。该算法考虑到数据集中各特征属性在聚类过程中不同的重要程度，重新构造关联数据点、聚类中心点的能量函数，利用弹性网络算法的求解模式，结合极大熵原理、模拟退火思想，提出一种具有加权特性的弹性网络聚类算法。该算法无需人工指导训练，便可以自学习地求解出高质量的聚类结果。通过不同维度、不同数量级的随机数据集和UCI真实数据集仿真实验，验证了算法的有效性和稳定性。相较于传统聚类算法，该算法显著提高了聚类质量。

Research on Clustering Algorithm of Elastic Net with Weighted Characteristics

Clustering analysis is one of the important contents in data mining. As a tool, clustering is important for people to analyze data. The weighted clustering direction of the elastic net is studied for the problems in clustering which include noise, high-dimensional and so on. This algorithm takes into account the importance of each feature in data set for the clustering process, reconstructs the energy function of the associated data points and cluster center points. Combining the maximum entropy principle, the idea of simulated annealing and the solution mode of elastic net method, a Weighting of Elastic Net for Clustering（WENC） algorithm is proposed. This algorithm can achieve high quality clustering results without manual training. Experimental results on both synthetic data sets and UCI real data sets show that WENC algorithm improves the quality of clustering.