基于遗传算法优化机器学习模型的地下水潜在性预测

Prediction of Groundwater Potential Based on Genetic Algorithm Optimized Machine Learning Model

传统机器学习模型在地下水潜在性预测中,未考虑最优因子组合,会对地下水潜在性制图产生不利影响。为此,提出了遗传算法优化支持向量机的地下水潜在性预测方法。以云南省彝良县为研究区,从地形、水文、土壤、地质等方面选取了共15个影响因子;考虑模型性能和影响因子的作用,利用遗传优化算法筛选了包含11个影响因子的最优因子组合;然后使用支持向量机方法构建了地下水潜在性预测模型;最后计算了因子优化前后的模型准确度和受试者工作特性曲线下面积(area under curve,AUC),并绘制了模型的受试者工作特性(receiver operating characteristic,ROC)曲线和地下水潜在性预测图。结果表明:因子优化前模型的准确度为0.774,验证集AUC为0.789,因子优化后模型的准确度为0.777,验证集AUC为0.806,分别提高了0.003和0.017。可见,所提方法的准确性、可靠性优于传统的支持向量机法,其结果可以为区域水文地质调查和地下水资源管理与规划提供科学参考。

Traditional machine learning models for groundwater potential prediction do not consider the optimal combination of factors,which can adversely affect groundwater potential mapping.For this reason,a groundwater potential prediction method based on genetic algorithm optimization support vector machine was proposed.Taking Yiliang County of Yunnan Province as the study area,a total of 15 influencing factors were selected from topography,hydrology,soil,geology,etc.Considering the model performance and the roles of the influencing factors,the optimal factor combinations containing 11 influencing factors were screened using the genetic optimization algorithm.Then the groundwater potential prediction model was constructed using the support vector machine method.Finally,the model accuracies and areas under the curves of the receiver operating characteristic(AUCs)before and after the optimization of the factors were calculated,and the receiver operating characteristic(ROC)curves of the model and the groundwater potential prediction maps were plotted.The results show that the accuracy of the model before factor optimization is 0.774,the AUC of the validation set is 0.789,the accuracy of the model after factor optimization is 0.777,and the AUC of the validation set is 0.806,which are increased by 0.003 and 0.017 respectively.It can be seen that the accuracy and reliability of the proposed method are superior to the traditional support vector machine method,and the results can provide scientific reference for regional hydrogeological survey and groundwater resource management and planning.