ALGORITMA GENETIK PENDUGAAN PARAMETER MODEL NONLINEAR JERAPAN FOSFOR (Genetic Algorithm for Parameter Estimation of Phosphorus Adsorption Nonlinear Model)

Abstract

Expectation Maximization (EM) was the best method used to estimate the parameters of phosphorus adsorption
in  a  nonlinear  model.    However,  it  is questionable  whether  the  optimum  value  obtained  was  exactly  a  global
optimum  value.    Genetic  algorithm  is  an  alternative  procedure  to  estimate  the  phosphorus  adsorption’s
parameters  in  a  nonlinear  model.    The  objective  of  this  study  was  to  have  a  better  understanding  in  the  use  of
genetic algorithm in maximum likelihood estimation of phosphorus adsorption nonlinear model parameters and
compare  it  with  the  EM  algorithm.   This  study  used   data  of  P  adsorption  isotherms  of  kaolinitic  and  smektitic soil  in  three  locations.   Phosphorus  adsorption  nonlinear models  used  are  Freundlich  and  Langmuir. Results showed that the genetic algorithm and EM method produced different values of estimated phosphorus maximum adsorption  and  bonding  energy  parameters.   AIC  and  SBC  values  of  genetic  algorithm  is  lower  than  EM algorithm, both on the Langmuir and Freundlich models.  AIC and SBC values of Langmuir model is lower than Freundlich  model  both  for  genetic  algorithms  and  EM  algorithm. Hence,  the  best  model  for  phosphorus adsorption is Langmuir nonlinear model  with genetic algorithm.
 
Keywords: nonlinear model, EM, Freundlich, Langmuir, genetic algorithm