改进的KNN-SMOreg算法及在铀矿床典型蚀变矿物赤铁矿含量预测中的应用

Improved KNN-SMOreg Algorithm and Its Application in Predicting the Amount of Hematite from Uranium

赤铁矿作为铀矿床蚀变带中一种典型的蚀变矿物,在铀矿勘测过程中具有十分重要的指矿作用.传统的建模方法都以同样的概率来处理每个特征参数的重要性,但在实际建模过程中各个参数的重要性是有差别的,忽略此因素会直接影响赤铁矿预测的精度.针对此问题,本文提出一种基于属性关联的加权K近邻和支持向量机的新混合模型WKNN-SMOreg,该模型通过属性关联的方式来赋予赤铁矿各特征参数权值,以此来降低关联程度较低的特征参数在预测过程中所引起的误差.实验结果表明:与K近邻、支持向量机、简单K近邻与支持向量机的混合模型KNN-SMOreg的预测方法相比,本文提出的WKNN-SMOreg方法提高了赤铁矿预测结果的精度,降低了无用信息在预测过程中所造成的负面影响.这说明属性关联加权技术在蚀变矿物含量预测中具有一定的应用价值.

Hematite,as a typical alteration mineral,plays a very important role in uranium exploration.Traditional modeling method usually treats every feature with the same probability.However,this does not hold in many real world applications,which may also cause the reduction of the accuracy of prediction.We propose a novel method called WKNN-SMOreg,which weights the features according to the association of their attributes on the hybrid of KNN and SMOreg.In this way,the error caused by the features with lower association will be reduced.The experiment results show,compared with KNN,SVM and KNN-SMOreg,the novel method improves the accuracy of prediction,and reduces the negative impact of the noise,which also implies that the new method can be well applied in the prediction of alteration minerals.