Fuzzy c-mean (FCM) clustering and Genetic Algorithm capability in predicting saturated hydraulic conductivity

Soil saturated hydraulic conductivity (K_s) is one of the key parameters as a main input for many water transport models in environmental studies. Direct measuring of this parameter is laborious, time consuming and expensive. So indirect prediction techniques such as Fuzzy c-mean (FCM) clustering and Genetic Algorithm was used to predict K_s parameter from other easily available metadata. FCM algorithm was used to cluster data, after that a Fuzzy Inference System had been generated based on this clusters by 12 rules, 6 numbers of inputs and saturated hydraulic conductivity as output. The FIS was trained by seventy percent of database using Genetic Algorithm. Based on statistical indexes (Pearson correlation coefficient, Maximum Error, Root Mean Square Error, Modeling Efficiency and Coefficient of Determination), results showed that in most cases, estimated K_sWas close to the measured K_s. Therefore, the use of FCM and GA techniques for estimating K_s is recommended.