مقاله انگلیسی رایگان در مورد الگوریتم خوشه بندی همگانی مبتنی بر فاصله مینکوفسکی – اسپرینگر 2017

In recent years, with the rapid development of the processing industry, the internet, cloud computing, mobile communication and web of things, data clustering has become an important research field that involves data security, data analysis and data mining. For example, complex systems in industry accumulate data that have the characteristics of enormous quantity, continuous sampling, multiple sources, and sparse values[1−3]. Higher requirements have been proposed for data processing in real-time, while research on fast data processing technologies has encountered many challenges. Clustering is a very useful technique for mining large data sets because it divides the data into smaller groups that are easier to address. Many different clustering methods have been investigated, including hierarchical agglomerative clustering, graph partitioning, mixture densities, and spectral clustering[4, 5]. Most of the clustering methods focus on finding a single optimal or near-optimal clustering according to some specific clustering criterion[6−10]. Consensus clustering is an important extension of classical clustering[11, 12]. Consensus clustering is used to find a compromise that provides a trade-off among different clustering information about the same data set. However, as one of the effective methods, consensus clustering is less sensitive to noise and can integrate solutions from multiple distributed sources of data or attributes. It solves the problem of reconciling clustering information about the same data set that arises from different sources. Many different approaches have been developed to solve the consensus clustering problem over recent years[7]. In consensus clustering algorithms, pairwise similarity often does not reflect a good measure of similarity between data points[13−15]. Addressing a large number of dimensions and a large number of data items is problematic due to time complexity. The effectiveness of the clustering methods depends on the definition of the similarity distance. In addition, the selection of the similarity matrix and the determination of the number of clusters are very difficult problems, and they must be solved urgently. Motivated by the above discussion, in this paper, we propose a novel consensus clustering algorithm. Based on the Minkowski distance[16, 17], the proposed algorithm can determine the number of clusters automatically and obtain the clustering results. This approach is also less sensitive to noise and can integrate solutions from multiple-sample data or the attributes of data in the process industry.