کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
1181312 | 962924 | 2008 | 7 صفحه PDF | دانلود رایگان |
We improve a recently developed Replacement Method (RM) for the selection of an optimal set of molecular descriptors from a much greater pool of such regression variables. Our approach yields almost optimal results with a much smaller number of linear regressions than the full search. We test our method on four different experimental full data sets and four sub datasets. The resulting algorithm, which was named Enhanced Replacement Method (ERM), resembles a simulated annealing procedure and improves our RM, yielding models with better statistical parameters than the ones previously published. The number of linear regressions increases only to a small extent so that the new algorithm is still suitable for databases with as many as 63912 descriptors.
Journal: Chemometrics and Intelligent Laboratory Systems - Volume 92, Issue 2, 15 July 2008, Pages 138–144