›› 2010, Vol. 53 ›› Issue (4): 420-426.doi:

• 研究论文 • 上一篇    下一篇

改进支持向量机在棉铃虫人工饲料配方优化中的应用

李俊, 谭显胜, 谭泗桥, 袁哲明, 熊兴耀   

  • 出版日期:2010-07-16 发布日期:2010-04-20
  • 通讯作者: 袁哲明

Application of improved support vector machine in the optimization of artificial diet for the cotton bollworm, Helicoverpa armigera (Lepidoptera: Noctuidae)

LI Jun, TAN Xian-Sheng, TAN Si-Qiao, YUAN Zhe-Ming, XIONG Xing-Yao   

  • Online:2010-07-16 Published:2010-04-20

摘要: 发展新的实验设计与分析方法,通过实施尽可能少的实验而获得满意配方对动植物营养、发酵工程等复杂多因素多水平寻优问题极为重要。本研究结合均匀设计(uniform design, UD)与支持向量回归(support vector regression, SVR),提出了一种新的配方优化实验设计与分析方法UD-SVR,将其应用于棉铃虫Helicoverpa armigera (Hübner)幼虫人工饲料配方优化。结果表明:在考虑6因素时仅通过2轮22个实验,表征配方优劣的指标——平均蛹重即由初始的0.2436 g高效提升至0.3044 g,明显优于二次多项式偏最小二乘回归等经验风险最小参比模型。UD-SVR预测精度高、指导性强、可解释性好、优化高效,有望在多因素多水平配方优化中得到广泛应用。

关键词: 棉铃虫, 人工饲料, 均匀设计, 支持向量回归, 配方优化

Abstract: The development of new experimental design and analysis methods that can provide satisfactory formula through conducting as few experiments as possible is extremely important for complex multi-factor and multi-level optimization problem, such as animal and plant nutrition, and fermentation engineering. Based on uniform design and support vector regression, a novel experimental design and analysis method named as UD-SVR for the prescription optimization was proposed. It was applied to optimize the complicated artificial diet including six factors for the cotton bollworm, Helicoverpa armigera (Hübner). The optimization results showed that the mean pupal weight increased from 0.2436 g in the initial benchmark formulation to 0.3044 g in the optimal prescription after carrying out only 22 schemes. UD-SVR is more efficient than the reference models and has the potential to be widely used for experimental design and analysis of the prescription optimization.

Key words: Helicoverpa armigera, artificial diet, uniform design, support vector regression, prescription optimization