遗传算法对5种理论生长方程的最优拟合研究

Optimal Fitting Study on Applying Genetic Algorithm to Five Theoretical Growth Equations

利用北美林肯小区黄杉6个分期的数据,选择Logistic、Richards、Korf、Gompertz和Mitscherlich 5种生长方程建立胸径随年龄变化的生长模型。分别利用遗传算法和传统拟合方法确定参数,并对拟合效果进行对比分析。研究结果显示:遗传算法对于Richards和Korf方程的拟合精度要明显高于传统拟合方法,对于Logistic、Gompertz以及Mitscherlich方程的拟合精度几乎一致。通过将检验数据带入由遗传算法拟合的方程中,除拟合后不符合生物学意义的Korf方程外,发现由剩下4种方程计算得到的预测值与实际值无明显差异(p值均大于0.05)。最终表明遗传算法对生长方程的最优拟合较传统拟合方法更有优势。

Based on the data of douglas fir from Lincoln Area in North America, five growth equation, such as Logistic, Rich- ards, Korf, Gompertz, and Mitscherlich, were utilized to establish the growth model in which DBH changes with age. Genetic algo- rithm and traditional fitting method were applied to determine parameters of the growth equation, respectively, and comparative analy- ses were exerted to the fitting effects between two approaches. The results showed that the fittiilg precisions of Richards and Korf by u- tilizing genetic algorithm was conspicuous superior to usual way, meanwhile, the fitting accuracies of Logistic, Gompertz and Mitscherlich were almost the same by exerting either of two techniques. By inputting inspection data into the fitting equation fitted through genetic algorithm, except for the Korf which is not corresponding with biological significance, difference between predictions calculated by genetic algorithm and actual values was extremely insignificant （ All values of p are above 0. 05 ）. It is clear that apply- ing genetic algorithm to optimal fitting on growth equation is more advantageous than the ordinary method ultimately.