Data mining predictive algorithms for estimating soil water content

Soil water content (SWC) plays a key role in the management of water and soil resources. Accurate prediction of SWC is an important issue in water and soil studies. Recently, some data mining and machine learning techniques were proposed for SWC prediction and achieved encouraging results. This paper presents four data mining predictive algorithms for SWC estimation. The used algorithms are random subspace ensemble, random tree (RT), reduced error-pruning tree (REPTree), and M5P. The motivation of this research is to investigate the performance of the popular data mining algorithms for SWC prediction. A benchmark dataset containing daily SWC parameters in three soil layers of 25 cm, 50 cm, and 100 cm from the Nebraska state station (central USA), Grand Island was used to evaluate the proposed techniques. Statistical indices of determination coefficient (R2), root-mean-square error (RMSE), mean absolute error (MAE), root relative square error (RRMSE), and relative absolute error (RAE) were utilized to measure the performance of the proposed prediction techniques. The modeling results showed that the RT algorithm with R2 = 0.97, RMSE = 0.38, MAE = 0.10, RRMSE = 7.32%, and RAE = 1.82% outperformed counterpart techniques. This study concluded that the developed models will help agricultural water users, developers, and decision-makers for achieving agricultural sustainability.

Automated summary

Soil water content (SWC) is crucial for water and soil resource management, and accurate prediction of SWC is important in water and soil studies. This research investigates the performance of four data mining algorithms for SWC prediction and finds that the random tree algorithm outperforms others. The developed models in this study can assist agricultural water users, developers, and decision-makers in achieving agricultural sustainability.