A graphical method - genetic algorithm (GMGA) approach for parameter estimation of twofold Weibull mixture model

This study presents a graphical method - genetic algorithm (GMGA) approach which combines graphical method (GM) and genetic algorithm (GA) for the parameter estimation of twofold Weibull mixture model. GM is firstly conducted to obtain crude estimates based on which the searching domain of each parameter is constructed to prevent the subsequent GA process from local convergence. Then GA searches the optimal parameter estimates according to the fitness function, which is a measurement of the absolute residuals between the fitted and the empirical distribution functions. The fitness function is developed through modifying the Kolmogorov-Smirnov (K-S) test so that it can also be directly used as a criterion of goodness-of-fit test. A numerical example is provided to verify the effectiveness of GMGA. The results show that the parameter estimates obtained by GMGA are more accurate than by some existing approaches.