基于PSO-SVM与样本加权的武器装备费用建模与预测  

The Modeling and Forecasting of Weapon Equipment Expenses Based on PSO-SVM and Sample Weighting

作　　者：袁冬根[1] 刘晓东[1] 王晓明[1] 蔡磊[1] 

机构地区：[1]空军工程大学工程学院,西安710038

出　　处：《火力与指挥控制》2011年第6期171-175,共5页Fire Control & Command Control

摘　　要：武器装备费用预测是武器装备费用分析的重要内容,预测分析的难点之一在于样本数据少,且样本数据具有复杂的非线性特点。充分利用支持向量机的结构风险最小化与粒子群算法快速全局优化的特点,采用粒子群算法快速优化支持向量机的模型参数;充分利用样本信息,模型中样本加权值的确定采用预测误差和样本相似度的样本加权方法,研究建立基于PSO SVM与样本加权方法的武器装备费用预测模型,进一步提高模型预测效果。最后,通过实例验证了该方法的可行性,为武器装备费用预测提供了一种新思路。Forecasting of weapon equipment expenses is the important content in expenses analysis of weapon equipment. One of the difficulties in forecasting and analyzing is the shortage and complicated nonlinear characteristics of sample data. This paper makes better use of the characteristics which is owed by the structure risk minimization of support vector machine and fast overall optimizing of particle swarm. The particle swarm optimization is used to optimize the parameters of the support vector machine. The sample weighting value in the model adopts the methods of estimate error and similarity of sample to make better use of the sample information. The prediction Model for Weapon equipment expenses based on PSO-SVM and sample weighting is established and improves the effect of model prediction. Finally, the feasibility of that method is proved with the example. The model provides a new thought for the research of weapon equipment expenses forecasting.

关 键 词：费用预测 粒子群算法 支持向量机 样本加权 

分 类 号：TN953[电子电信—信号与信息处理]