Prediction of cumulative infiltration of sandy soil using random forest approach. Issue 2 (3rd April 2019)

• Record Type: Journal Article
• Authors: Sihag, Parveen; Tiwari, N. K.; Ranjan, Subodh
• Is Part Of: Journal of applied water engineering and research. Volume 7:Issue 2(2019)
• Page Start: 118
• Page End: 142
• Publication Date: 2019-04-03
• Subjects: cumulative infiltration -- random forest -- M5P tree regression -- bagged approach -- cumulative time; Water-supply engineering -- Periodicals; Water-supply engineering; Periodicals; 627.05
• DOI: 10.1080/23249676.2018.1497557
• Languages: English
• ISSNs: 2324-9676

Abstract

Abstract : This paper aims to examine the performance of soft computing models (bagged and unbagged with Random Forest (RF) and M5P tree regression models) in the estimation of cumulative infiltration. Performances of these soft computing techniques were compared with previous studies of cumulative infiltration of soil. Laboratory experiments were carried out on soil samples with predetermined moisture contents and different compositions of rice husk ash and fly ash and accordingly, 413 observations were obtained. The evaluation of results suggests that the RF model performs better than other considered models and it could effectively be used in the modeling of the cumulative infiltration. The bagged approach was found to perform well with the M5P tree model than the RF model. Sensitivity analysis concludes that cumulative time, suction head and moisture content were the most important parameters. In addition, parametric studies were also carried out.