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

Soil saturated hydraulic conductivity (Ks) is one ofthe key parameters as a main input for many water transportmodels in environmental studies. Direct measuring of thisparameter is laborious, time consuming and expensive. Soindirect prediction techniques such as Fuzzy c-mean (FCM)clustering and Genetic Algorithm was used to predict Ksparameter from other easily available metadata. FCM algorithmwas used to cluster data, after that a Fuzzy Inference Systemhad been generated based on this clusters by 12 rules, 6 numbersof inputs and saturated hydraulic conductivity as output. TheFIS was trained by seventy percent of database using GeneticAlgorithm. Based on statistical indexes (Pearson correlationcoefficient, Maximum Error, Root Mean Square Error,Modeling Efficiency and Coefficient of Determination), resultsshowed that in most cases, estimated Ks was close to themeasured Ks. Therefore, the use of FCM and GA techniques forestimating Ks is recommended.

Keywords:

Soil saturated hydraulic conductivity , FuzzyInference System , Fuzzy c-mean Algorithm , Genetic Algorithm