主成分-距离水质识别水源模型主成分的选取

Select principal components in principal component-distance water source identification model by quality

为了寻找最佳主成分以使水质识别水源模型效果最好,以焦作矿区水质数据为例建立主成分-距离水源识别模型并进行回代验证和交叉验证,正确识别率达到82.9%;同时对四个不同矿区分别建立模型,发现主成分分别选取4、4、4、3个时识别效果最好,回代正确率和交叉正确率分别达到87.5%、97.1%、93.9%、93.1%和83.3%、94.3%、90.9%,89.7%。结果表明:主成分-距离水质识别水源模型能够很好地判别煤矿突水来源;在交叉验证率最大时,选取最大回代验证率对应的主成分,模型识别效果最好。

To find the best principal component to make the best effect of water quality identification model,Taking Jiaozuo water quality data as an example,the principal component distance water source identification model was established,and the correct recognition rate was 82.9 %;at the same time,the model was established for four different mining areas,and it was found that the recognition effect was the best when the principal component was selected as 4,4,4,3,respectively,and the correct rate and intersection of the back generation were achieved The correct rate of fork was 87.5 %,97.1 %,93.9 %,93.1 % and 83.3 %,94.3 %,90.9 % and 89.7 % respectively.The results show that the principal component distance water quality identification water source model can identify the source of coal mine water inrush well;when the cross validation rate is maximum,the principal component corresponding to the maximum back validation rate is selected,and the model identification effect is the best.