基于战例数据的制胜要素选择方法

针对战争制胜机理定量研究需求,提出了一种基于混合评估的制胜要素选择算法。针对战例数据特点,该算法选择两种过滤方法分别从不同方面对要素全集进行评估排序而后加权得到综合排序结果;将结果作为遗传算法的初始种群,而后以分类精度作为个体适应度函数。选择几种典型的分类器综合比较,筛选出规模较小、性能较好的要素子集。测试结果表明,该算法不仅能有效地减少要素子集规模,还可以进一步提高制胜机理分析模型的准确率和效率。

Research on Method of Winning Factors Selection Based on Data of Battle Cases

For the Demand of factor selection in quantitative study of winning mechanism,and considering the inadequaces of traditional methods,a new factor selection model based on filter-wrapper is proposed to select the decisive factors in war. The model combines two filters to pre-rank all the variables in the battle cases dataset from different aspects,and then produce an initial GA population based on it. In the wrapper,Genetic Algorithm is selected to search the factor subsets. In GA,individual fitness degree is evaluated by classification accuracy of multi-classifier,which can help find the subsets with both of smaller size and better performance. Tests demonstrate that the FSFW model not only can reduce dimensionality of factor subset,but also can improve the accuracy and efficiency of winning models.