鉴于变量选择在QSAR/QSPR研究中的重要性，比较了遗传算法和几种传统的方法，如前进法、后退法及逐步回归法.结果表明，对于研究中所用数据，遗传算法较几种传统的方法为好，其原因可能由于传统的方法陷入了局部最优.遗传算法在变量较多的情况下方可显示出效率高和得到较好结果的优越性.对于变量的选择，遗传算法是一值得推荐的有效的方法.
Since variable selection is important in quantitativestructure-property/activity studies, in this paper comparison between the genetic algorithm and several common methods such as forward method, backward elimination and stepwise regression is performed. The results obtained by using the common methods are very close to each other and that obtained by using genetic algorithm are better than the former methods. Though genetic algorithm takes time than the common methods, it is more effective for variable selection, especially in the case that there are a lot of descriptors to be treated.
参考文献
| [1] | XU Lu(许禄)Auth(著). Methods of Chemome-trics(化学计量学方法). Beijing(北京):Science Press(科学出版社),1995 |
| [2] | Holland J H Auth. Adaptation in Natural and Artificial Systems. Ann Arbor,University of Michigan Press,1975 |
| [3] | Hibbert D B. Chemom Intell Lab Syst,1993,19:277 |
| [4] | Hasegawa K,Miyachita Y,Funatsu K. J Chem Inf Comput Sci,1997,37:306 |
| [5] | Hasegawa K,Funatsu K. J Mol Struct(THE-OCHEM),1998,425:255 |
| [6] | Norinder U. J Chemom,1996,10:95 |
| [7] | ZHANG Yuan(章元),ZHU Er-Yi(朱尔一),ZHUANG Zhi-Xia(庄峙厦), et al. Gaodeng Xuexiao Huaxue Xuebao(高等学校化学学报),1999,20(9):1371,1375 |
| [8] | HE Xian-Feng(何险峰),ZHOU Jia-Ju(周家驹). Jishuanji Yu Yingyong Huaxue(计算机与应用化学),1998,15(4):211,216 |
| [9] | Hall L H. Environ Toxico Chem,1989,18:431 |
- 下载量()
- 访问量()
- 您的评分：
-
10%
-
20%
-
30%
-
40%
-
50%