基于RAGA的投影寻踪分类模型改进与实例分析

Based on Projection Pursuit Classification Model Improvement and Analysis of Examples RAGA

针对实数编码加速遗传算法(RAGA)在求解投影寻踪分类(PPC)模型陷入局部最优的问题,通过引入区间扩展因子:在变量区间过小时,对变量区间进行适当扩展;在扩展区间"越界"时,即以边界作为变量的取值。并选取合理的局部密度窗口半径R,建立了改进的RAGA-PPC分类模型,并以文献中S县15个乡镇申请粮援项目的投资顺序为例进行验证分析。研究表明,改进的RAGA-PPC模型对样本分类评价,确立指标因素的贡献程度大小具有一定的可行性和广泛的通用性。

Aiming at the problem that real-coded accelerating genetic algorithm （RAGA） could not solve global optimal solution of PPC Model, this paper proposes an improvement ： when variableinterval is too small, then extends variable interval by an appropriate constant; when the extension crosses the border, set the boundary as the variable＇s value. Combining with proper R value, the improved RAGA-PPC model is established, and using it in food aid pro- ject investment order of S county＇s 15 towns, more reasonable sequences and each factor＇s influence on the investment order are obtained, The results show that the improved RAGA-PPC model has strong applicability and generality of sample classification and evaluation as well as estimating each factor＇s contribution.