摘要
分类技术是数据挖掘的重要分支,常见的分类方法有决策树、统计方法、机器学习方法、BP神经网络方法等。本文针对标准BP网络存在的一些缺陷,结合一种进化算法微粒群(PSO)算法,建立了一种用于数据分类的网络模型。该模型充分利用微粒群算法的全局寻优特性,优化BP网络的权值和阈值,既保证了BP网络能收敛到全局最优解,加快了BP网络的收敛速度和收敛精度,又提高了待分类数据的识别准确率。仿真实验结果表明此模型较BP网络具有较好的分类识别性能。
The classification technology is an important branch of data mining. This paper introduces the basic ideas of particle swarm optimization algorithm. A network model is established for data classification, which is composed with PSO algorithm and BP network. The model takes full advantages of global searching of PSO. It optimizes the weights and biases of BP network and also accelerates the network' s convergence. The computer simulation result shows that it has preferable classification and recognition capability
出处
《自动化技术与应用》
2007年第11期13-15,18,共4页
Techniques of Automation and Applications
基金
内蒙古自然科学基金(200408020809)
关键词
微粒群算法
BP网络
模式识别
数据挖掘
particle swarm optimization
BP network: pattern recognition
data mining