基于多类SVM武器装备预研项目风险预警模型及实证研究

在分析武器装备预研项目研究范畴及特点的基础上,进行全寿命周期内项目风险因素识别。对现有风险预警模型进行了比较分析,针对现有研究模型的不足,将支持向量机（Support Vector Machine, SVM）算法应用于武器装备预研项目风险预警,选取1-a-1方法对SVM进行多类扩展,选择RBF函数作为核函数,构建基于多类SVM的武器装备预研项目风险预警模型。最后,结合某军工研究所预研项目进行了实证研究,结果表明该方法拟合精度高、鲁棒性强,对于同类问题具有一定的推广价值。

Risk Warning Model of Weapon Development Projects Based on Multi-class SVM and Its Empirical Study

On the basis of research scope and characteristics of weapon development projects, this paper identified project risk factors in the full life cycle. A comparative analysis of the existing risk early warning models was given. For the lack of existing research model, this paper applied the SVM method to weapon development projects risk warning, which selects 1-a-1 method for SVM multi-class extensions, and uses the RBF function as the kernel function. Finally, this paper took a military institute to make empirical research. The results showed that the method was fitting accuracy, robustness, and had some promotional value for the same sort of problems.