试验设计支持向量机方法在新药试生产中的应用.PDFVIP

第25卷第8期 计算机应用 V0l_25 No．8 2005年8月 Computer Applications Aug．2005 文章编号：1001—9081(2005)08—1919—04 试验设计、支持向量机方法在新药试生产中的应用 朱娟娟，张申生，杜 涛 (上海交通大学计算机系，上海200030) (zhujuanjuan@sjtu．edu．CB) 摘 要：传统的新药试生产过程中，试验方案设计、数据分析大多是人为控制的，主观性太强。针 对这些情况，建立了一套采用试验设计、支持向量机方法进行新药试生产的系统。首先采用试验设计 方法科学合理地生成试验方案，并进行试验。对于试验数据，采用适合小样本数据进行分析建模的支 持向量机(SVM)方法来建立回归模型，并用模型进行试验结果预测和方案优化。同时简单介绍了基 于贪婪搜索的交互式优化方法等。最后以一个实例验证了该套方法的科学有效性。 关键词：正交试验设计；均匀试验设计；SVM方法；贪婪搜索 中图分类号：TP311．52 文献标识码：A Applying experiment-design and support-vector—machines methods to new medicines’prepr0ducti0n ZHU Juan-juan，ZHANG Shen·sheng，DU Tao (Department ofCompute Science，Shanghai Jiao University,Shanghai 200030，China) Abstract：In the traditional process of new-medicines’preproduction，designing experiment schemes and analyzing data was usually manipulated artificially．It was tOO subjective．Aiming at these situations，a new·medicines’preproduction system was presented，which was based on experiment-design and support—vector-machines methods．First，experiment-design method Was introduced in order to gain scientific and logical experiment schemes．Then，support-vector-machines(SVM)method Was adopted to estabKsh regress models for those experimental data，which is an analyzing and modeling tOOl suiting small sample data．These models were used for predicting the experiments’ results and optimizing the schemes． In addition， an altemant optimizing method based on greedy search was introduced too．Finally，an exam ple Was given to validate that the whole set of methods Was scientific and effective． Key words：orthogonal experiment design；unif