摘要
针对现有的Neural-Gas算法进行改进,提出了一种新的聚类算法。改进之处在于:一个点对一个簇的质心的影响程度取决于该点到其他更近的簇的质心的距离值,而不仅仅是点与簇质心间距离值按大小排列次序的序号。在几个数据集上的实验结果表明,该算法在熵、纯度、F1值、rand index、规范化互信息NMI等五个指标上优于K-means算法、Neural-Gas算法等其他几种聚类算法,该算法是一种较好较快的算法。
This paper proposed a new clustering algorithm by improving existed algorithm Neural-Gas. The improvement was that the degree of influence of a point on a cluster centroid depended on the distance values between this point and the other more near cluster centroids,not just the sequence number arranged in the order of the distance value between this point and cluster centroids. Experimental results on several data sets show that on five indicators such as entropy,purity,F1 value,rand index and normalized mutual information this improved algorithm surpasses K-means,Neural-Gas and several other algorithms.The conclusion is that this improved algorithm is a better and faster algorithm.
出处
《计算机应用研究》
CSCD
北大核心
2015年第5期1309-1312,共4页
Application Research of Computers
基金
国家自然科学基金青年资助项目(71102149)