摘要
根据免疫算法的基本思想 ,提出了一种基于免疫识别 ( IR)的主因子数判断方法 .对模拟数据和HPLC-DAD数据的处理结果表明 ,IR比嵌入误差 ( IE)、因子指示函数 ( IND)和交叉验证 ( CV)更能有效地克服噪音的影响 .当实际数据中的非线性和不等性方差噪音使其它方法失效时 。
Based on the idea of the immune algorithm, an immune recognition(IR) method was proposed to obtain the principal factor number of a data matrix. The original matrix is input to the algorithm as antigens, a number of antibodies are generated by means of PCA, elimination of antigens is performed in a recurrence procedure. An equation for the calculation of the affinity value is defined. In each elimination, an affinity value is calculated. According to the series of the calculated affinity values, the principal factor number of the matrix can be obtained. By the tests of both simulated and experimental HPLC DAD data, the IR method showed the superiority over the imbedded error(IE), the factor indicator function(IND) and the cross validation(CV). In treating experimental HPLC DAD data, due to the interference of the nonlinearity and the heteroscedastic noise, only IR can obtain the correct factor number, none of IE, IND or CV succeeded.
出处
《高等学校化学学报》
SCIE
EI
CAS
CSCD
北大核心
2003年第4期595-598,共4页
Chemical Journal of Chinese Universities
基金
国家自然科学基金 (批准号 :2 9975 0 2 7)
