摘要
研究比较了最小二乘支持向量机(LSSVM)、BP神经网络(BPNN)对苯酚、对苯二酚和间苯二酚混合液的紫外分光光度法同时测定,测定结果各组分平均相对误差LSSVM分别为2.75%、2.43%、1.45%;BPNN分别为2.97%、3.35%、4.72%;LSSVM模型测出量的相对误差的绝对值最大为5.73%,BPNN为16.10%.结果表明LSS-VM用于该混合体系的同时测定数据解析,处理结果的准确性明显优于BPNN方法.
We have studied and comparde direct quantitative determination of phenol, m- , p- dihydroxybenzene by UV spectroscopy with the help of LSSVM and BPNN, The mean relative error of prediction results of synthetic samples for LSSVM and BPNN：LSSVM is 2.75%, 2.43%, 1.45% ,BPNN is 2.97%, 3.35%, 4.72% ;the maximal absolute relative error of LSSVM is only 5.73% ,BPNN is 16. 10%. The resuhs show that LSSVM model is much better than BPNN in accuracy.
出处
《西华师范大学学报（自然科学版）》
2006年第2期217-220,共4页
Journal of China West Normal University(Natural Sciences)
