Document Type : Research Paper

Authors

• Maryam AriaSadr ¹
• Dariush Rahimi ²
• Mehran Zand ³
• Hadi Amiri ⁴

¹ Physical Geography ,Iniversity of Isfahan

² Physical Geography ,University of Isfahan

³ soil conservation and watershed management research institude, agricultural research, education and extension.organization (AREEO),

⁴ Department of Economics, Faculty of Economics and Administrative Sciences, University of Isfahan, Isfahan, Iran

Abstract

Water quality is an important indicator of health. Determining water quality requires expensive tests. In this research, Support Vector Machine (SVM) and Classification And Regression Tree (CART) algorithms have been evaluated. To calibrate the techniques, 321 test samples of physical-chemical elements of Cham Anjir station (1969-2021) were used. According to the results of the correlation matrix, Total Hardness (TH) in water and Total Dissolved Solids (TDS) in water were selected to evaluate and select the optimal model. The results of the trend test showed that the concentration of both indicators has increased since 1985. The evaluation indicators of the SVM, showed that the best results of the SVM were obtained in the KernalLinear model and in the CHAID method (CART) algorithm at the 95% combination threshold. 10 rules were presented for estimating TDS, and 119 rules for estimating TH . The Validation of SVM and CART algorithms of observational and estimated data showed that MSE, MAE, R2 coefficients are not different in both SVM and CART models. But the estimations of TH in water and TDS. The amount of RMSE in SVM was less than 10%. Therefore, the SVM obtained better results compared to the CART for estimated data.

Keywords

• Decision Tree
• Support Vector Machine
• Total Hardness
• Validation
• Water quality