多元宇宙优化算法应用于聚类分析

Multi-verse Optimization Algorithm for Cluster Analysis

关键词:  数据挖掘  k-均值  聚类问题  多元宇宙算法

Key words:  data mining  k-means  clustering problems  multi-verse optimization algorithm

摘要:

[目的]聚类是数据分析和数据挖掘技术中最重要的概念,其中,k-均值聚类算法是最常用的方法之一。然而,k-均值聚类算法高度依赖于初值,容易陷入局部最优解。为了克服k-均值聚类算法存在的不足,[方法]本研究提出一种利用多元宇宙算法(MVO)解决聚类分析问题的新算法,并进行一些数据集测试实验。[结果]数值模拟实验表明多元宇宙算法解决聚类问题效果优于人工蜂群(ABC)算法,布谷鸟搜索(CS)算法、粒子群优化(PSO)算法等。[结论]在大多数测试数据集的情况下多元宇宙算法解决聚类分析问题具有收敛速度快、聚类精度高和稳定性好的优点。

Abstract:

[Objective] Clustering is a popular data analysis and data mining technique. The k-means clustering algorithm is one of the most commonly used methods. However the k-means clustering algorithm is highly dependent on the initial solution and it is easy to fall into local optimal solutions.[Methods] For overcoming these disadvantages of the k-means method, this paper proposed Multi-verse Optimization Algorithm for the cluster analysis and experiment on synthetic and real life data sets. The numerical simulation experiments and comparisons were carried out based on a set of test data. The MVO algorithm was compared with Artificial Bee Colony (ABC) algorithm, Cuckoo Search (CS), Particle Swarm Optimization (PSO).[Results] From the experimental result, we could discover that MVO performed the best in most data set cases. We could easily find that the convergence speed of MVO was faster than other algorithms mentioned in this paper in most cases. Another fact could be found that the stability of MVO could reach a relatively high level as well.[Conclusion] Both numerical experiment results and the graphical experiment results show that Multi-verse Optimization Algorithm is more competitive than other algorithms for solving the clustering problem.