摘要
针对基本粒子群算法易陷入局部解的不足,本文基于基本粒子群算法,着重对惯性权重因子进行改进,在非线性递减惯性权重策略基础上增加随机因素的考虑,给出了改进的算法—非线性递减随机惯性权重粒子群算法。并利用国际常用基准测试函数进行仿真实验,测试结果验证了改进算法的计算性能优于基本粒子群算法。在此前提下,本文针对多元线性回归分析中的参数计算复杂问题,又提出一种基于上述改进算法的参数估计方法,以最大似然准则作为粒子群优化算法的适应度函数,建立多元线性回归分析中的参数估计计算模型。算例仿真结果显示,该方法是高效和实用的。
As for the flaw of standard PSO getting into early local optimum more easily, an improved PSO（Non-Linear Decreasing Random Inertia Weight PSO）is proposed based on modifying the inertia weight of standard PSO. It is a new strategy of inertia weight to add the considering of random factors based on Non-Linear Decreasing Inertia Weight. Experiments on the benchmark functions show that the performance of the improved PSO outperforms standard PSO. In order to solve the complex problems of the parameter estimation calculation for the multiple linear regression models, a novel method to estimate parameters is presented based on the improved particle swarm optimization algorithm in the paper. Maximum likelihood estimation is adopted as the fitness function for the optimization problem. Thus the model of calculating parameters for the multiple linear regression models is set up. Through a numerical simulation computational experiment, the effectivity and practicality of this method is demonstrated.
出处
《计算机工程与科学》
CSCD
北大核心
2010年第4期101-105,共5页
Computer Engineering & Science
基金
浙江省2010年度教育科学规划研究课题(SCG194)
关键词
粒子群优化算法
惯性权重
参数估计
多元线性回归
particle swarm optimization
inertia weight
parameter estimation
multiple linear regression models
