无参数自然邻近聚类算法CB3N：性能优于传统方法

61 浏览量更新于2024-08-27 1 收藏 373KB PDF 举报

本文主要探讨了一种新颖的基于自然最近邻（Natural Nearest Neighbor, NN）的聚类算法——CB3N（Cluster Based on Natural Nearest Neighbor）。该算法在《计算机信息系统》杂志2014年第10期第13卷发表，作者是Qingsheng Zhu、Jinlong Huang、Ji Feng 和 Xianlin Zhou，来自重庆大学计算机科学学院的软件理论与技术重点实验室。传统的聚类算法在数据分析、模式识别和图像处理等领域广泛应用，但它们在处理数据集时往往面临一个挑战：如何在没有预先设定参数的情况下选择合适的划分方式。CB3N算法正是针对这个问题提出的，它摒弃了对参数的依赖，通过采用一种无参数的搜索方法来确定每个数据点的自然邻居。自然最近邻是一种新的邻域概念，它不依赖于任何预先设定的距离阈值或相似度度量，而是根据数据点之间的内在结构进行自动判断。 CB3N的核心思想在于通过寻找每个数据点在数据集中的自然邻居，这些邻居具有与该点相似的内在属性或者结构特征，而非仅仅是最直观的欧氏距离最近的点。这种无参数的搜索策略使得CB3N算法具有更高的适应性，能够在不同类型的复杂数据集上自适应地进行聚类。实验和性能分析部分展示了CB3N的优势。首先，CB3N在没有预先设定参数的情况下也能有效地对数据进行分组，这在实际应用中极大地简化了预处理步骤。其次，对比其他代表性的聚类算法，如K-means、DBSCAN等，CB3N在许多情况下能够获得更好的聚类结果，这表明其在保持简单性和高效性的同时，还能提高聚类的精度和有效性。关键词包括自然最近邻（Natural Nearest Neighbor）、聚类算法以及无参数方法，这些都是CB3N算法的关键组成部分。这篇研究论文为无参数聚类问题提供了一个创新且实用的解决方案，为未来在大数据和高维空间下进行无监督学习提供了新的思路和技术支撑。

Journal of Computational Information Systems 10: 13 (2014) 5473–5480

Available at http://www.Jofcis.com

A Clustering Algorithm Based on Natural Nearest

Neighbor

Qingsheng ZHU

∗

, Jinlong HUANG, Ji FENG, Xianlin ZHOU

Chongqing Key Lab. of Software Theory and Technology, College of Computer Science, Chongqing

University, Chongqing 400044, China

Abstract

The Clustering has been widely used in many areas such as data analysis, pattern recognition, image

processing. It is very diﬃcult for most well-known clustering algorithms to select an appropriate

parameter when they run on a dataset. In this paper we propose a novel clustering algorithm based on

natural nearest neighbor (CB3N). Natural Nearest Neighbor is a new concept of nearest neighbor which

adopts a parameter-less algorithm of searching the natural neighbors for each point in a dataset. Our

experiments and performance analysis demonstrate that CB3N could not only cluster a dataset without

any parameter, but also get better clustering results than other representative clustering algorithms.

Keywords: Natural Nearest Neighbor; Clustering Algorithm; Parameter-less

1 Introduction

Data Mining is to ﬁnd implicit, eﬀective, valuable and comprehensible mode from large amounts

of random data, then to discover useful knowledge. Clustering is a primary method of data

mining and data analysis. Clusters are collections of objects whose intra-class similarity is high

and inter-class similarity is low. The research on clustering algorithm has been very active

[1]. New clustering methods have been constantly proposed. They can be roughly divided into

Partitioning Methods, Hierarchical Clustering, Density-Based Clustering, Grid-Based Clustering,

Model-Based Method, and so on.

The concept and algorithm about k-nearest neighbor are foundation scientiﬁc issue in many

ﬁelds of application study. Thus the improved algorithms based on kNN and theirs applications

have being become the hot research ﬁelds such as data mining [2], pattern recognition [3], image

processing [4] and intelligent recommendation [5].

However, there are some technological restrains that all of these algorithms almost can’t eﬀec-

tively cluster without appropriate parameter. In other words, most of these algorithms have high

dependency to the parameter. Once the parameters changed, the result of clustering would have

obvious diﬀerence. So the selection of parameter is very important for clustering algorithm. In

Project supported by the National Nature Science Foundation of China (No. 61272194).

∗

Corresponding author.

Email address: qszhu@cqu.edu.cn (Qingsheng ZHU).

DOI: 10.12733/jcis10521

July 1, 2014

下载后可阅读完整内容，剩余7页未读，立即下载

weixin_38699492

粉丝: 8
资源: 946

无参数自然邻近聚类算法CB3N：性能优于传统方法

模式识别 最近邻规则聚类算法

一种基于邻居信息的派系过滤算法（论文+算法+代码）

聚类算法总结

密度峰值聚类算法改进

几千万个数据用Dscent聚类算法

基于轮廓系数的聚类评价算法

DBSCAN聚类算法缺点

KNN算法是分类算法还是聚类算法

DBSCAN聚类算法

java 编制最近邻设施算法 。

最新资源

模式识别最近邻规则聚类算法

java 编制最近邻设施算法。