基于Minkowski距离的一致聚类改进算法及应用研究

针对一致聚类算法中聚类数目判断不准确、聚类速度慢等问题,通过集成复杂网络中的Newman贪婪算法与谱聚类算法,提出了一种新的基于Minkowski距离的一致聚类算法.该算法利用Minkowski距离刻画样本间的相似度,根据随机游走策略,结合不同数据的特征值分布分析方法进行聚类,实现聚类数目的自动识别.实验仿真说明算法具有较少的运算时间及较高的聚类精度.结合实际铜矿泡沫浮选过程特点,将该算法应用于浮选工况分类,进一步验证了算法的有效性.

Research on Improved Consensus Clustering Algorithm Based on Minkowski Distance and Its Application

Aiming at the inaccuracy of clustering numbers and the slow speed of ordinary consensus clustering algorithms, Newman greedy algorithms of complex networks theory and spectral clustering algorithms were combined to propose a novel consensus clustering algorithm based on Minkowski distance. The algorithm depicts the similarity between samples in terms of Minkowski distance and adopts the strategy of random walk. By adjusting the parameters of the Laplacian distance, the accurate information of the clustering number is automatically obtained. The simulation results show that the proposed consensus clustering algorithm based on Minkowski distance has the superiority of the running time and accuracy of the clustering number. This method was applied to actual copper froth flotation process, and the results further illustrated its effectiveness.