Open AccessThe optimization of kink regression with differential evolution (DE)
Author(s) -
Pathairat Pastpipatkul,
Petchaluck Boonyakunakorn,
Songsak Sriboonchitta
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1324/1/012087
Subject(s) - estimator , maximum likelihood , monte carlo method , mathematics , statistics , regression , differential evolution , global optimization , linear regression , mathematical optimization , function (biology) , econometrics , computer science , evolutionary biology , biology
This paper aims to estimate the kink model with respect to Thailand’s exports to Thai GDP. Another main contribution of this study is to compare the performance of optimization the kink regression model with DE and MLE. The traditional optimization is maximum likelihood estimator (MLE) which has problems in the case where the function is nondifferentiable, or the likelihood is difficult to find. Furthermore, it is difficult to obtain the global maximum in the non-linear model. One of the optimization techniques is DE which is a successful method for searching for the global maximum. To evaluate the estimated performance of DE and MLE, we apply the Monte Carlo simulation. The simulation result indicates that DE has the both lower MSE and bias. We apply the regression model with respect to Thai exports to Thai GDP. For the estimated parameter results show that the kink point is -0.016%. The estimated coefficients in the first part and second part are 1.144% and 0.341%, respectively.