Prediction of groundwater level using the hybrid model combining wavelet transform and machine learning algorithms

The main objective of this study is to predict groundwater levels (GWLs) using modified hybrid algorithms with two levels of improvement. The observed GWLs, precipitation, and temperature were used as input variables in the prediction algorithms. Two widely used machine learning algorithms, namely support vector machine (SVM) and random forest (RF), were used first as the base algorithm, then the wavelet transform (WT), with seven wavelet types, was employed as a preprocessing method. SVM and RF combined WT, namely W-SVM and W-RF, respectively, constructed the first-level hybrid algorithm, taking all of the components together. In addition, the approximation and detail components were separately used to construct the second-level hybrid algorithm, coupling SVM and RF with WT, namely W-SVM-D and W-RF-D, respectively. Four statistical metrics were used to evaluate and validate the predictive accuracies of algorithms. According to the obtained results, The W-SVM-D and W-RF-D hybrid algorithms demonstrated the highest predictive accuracies of GWLs, followed by the first-level hybrid and single algorithms. In addition, the mother wavelet types affected the prediction accuracy of W-SVM and W-RF algorithms, while W-SVM-D and W-RF-D algorithms showed the highest predictive accuracies of GWLs for all the selected wavelets. It is suggested that the modified hybrid model can be effectively used to predict groundwater levels.

Automated summary

The main objective of this study is to predict groundwater levels using modified hybrid algorithms. The results showed that the W-SVM-D and W-RF-D hybrid algorithms had the highest predictive accuracies among all the selected wavelets.