基于支持向量机的舰船维修费用组合预测研究

在费用预测中,利用单一模型往往存在着信息不足的缺陷。为了提高舰船维修费用的预测精度和稳定性,采用支持向量机(SVM)回归算法,把几种单一预测模型结果作为输入,实际值作为输出,然后用足够多的预测案例训练学习机器,在各组合的模型预测结果与实际之间得到一种非线性映射关系,从而建立了非线性组合预测模型。最后,以某型舰船维修费用为例,对指数平滑法、灰色预测和参数法3种方法的预测结果进行仿真,结果表明此法较传统的单一模型预测法具有更高的预测精度。

In warship maintenance costs forcast, problems always exist for the single model because of information limited, In order to improve forecast precision and stability, support vector machine （SVM） regression arithmetic is adopted. A nonlinear mapping relation is found by inputting single forecast result, outputting the real value, and training the machine with adequate cases. Consequently, nonlinear combined forecast model based on SVM is built up. At last taking certain warship maintenance for example, we simulate the combined forecast process, which combines the forcast result of exponential smoothing approach, grey prediction and parametric method. Final result shows that this method is superior to traditional single model forecast in precision.