Research on the Similarity Algorithm of Chromatographic Fingerprint Based on Information Entropy

摘要

In order to reduce the error of interangle cosine method or correlation method in the chromatographic fingerprints analysis, an improved similarity algorithm that assigns different weight according to information entropy was proposed in this paper. In this way, 23 samples of Exocarpium Citrus Grandis collected from different places were obviously classified as two different species, which solved the problem that it is difficult to make an accurate identification either by the traditional similarity algorithm or its' optimization by coefficient of variation. Furthermore, better results were obtained by compared with discrimination model base on principal component analysis and artificial neural network. So, the method based on information entropy was more suitable for quick, effective discrimination of species and origin of traditional Chinese medicine.