The Study Of Boundary Detecting Algorithm With Clustering

The contemporary world is an information-oriented world, there are large amounts of data in all walks of life, it is an urgent need to convert these data into useful information and knowledge to help people do business management, production control, market analysis, engineering design and scientific exploration, etc. So data mining have emerged as the focus of attention in the IT industry. Data mining is to extract useful knowledge from a large number of data; it involves many disciplines, including database technology, artificial intelligence, machine learning, pattern recognition, statistical theory, information theory, high-performance computing and so on. Cluster analysis is an important research topic in data mining research field and has been applied to pattern recognition, image processing, data analysis, market research and many other fields. At the same time, cluster boundary analysis also has a high research value in data mining, and has been widely used in biology, physics, image recognition and other areas. Now there have been many clustering algorithms and boundary detection algorithm, but most of these algorithms are independent of each other, and do not integrate clustering and boundary detection phase. In addition, previous boundary detecting algorithms have many shortcomings in practical applications and it is difficult to achieve the user's desired effect.This dissertation summarizes the results of previous studies, and depthly analyse boundary detection algorithm existed and find that these algorithms are either inefficient, or boundary detection accuracy is low, or the input parameters difficult to determine. To overcome these shortcomings, this paper present the conception of entropy of angle between vectors and a boundary points detecting algorithm called BDVE(a Boundary Points Detection algorithm based on Entropy of Angle between Vectors).The algorithm makes use of the advantages of grid which can improve the computing speed and effectively remove the noise. For the problem that present/current clustering algorithms and boundary detection algorithms are separated, this paper proposes an algorithm named cluster boundary detecting algorithm based on Delaunay triangulation (DTBOUND). It integrates clustering and boundary detection, and makes full use of the benefits that the coefficient of variation and delaunay triangulation graph can naturally reflect the distribution of data points.This dissertation implements the algorithm BDVE and DTBOUND, and does a lot of experiments in the integrated data sets, and compare with other boundary detection algorithms. Experimental results show that both BDVE and DTBOUND can quickly and effectively identify cluster boundary points of clusters of arbitrary shapes, different sizes and different densities.At the same time, DTBOUND can quickly and effectively identify clusters of arbitrary shapes, different sizes and different densities.