Detection and prediction of active ingredients in fingerprints of traditional Chinese medicine from a high-dimensional perspective

High-dimensional problems exist in various fields of medicine. Data analysis of medical data from a high-dimensional perspective brings new challenges for researchers. This paper intends to analyze the fingerprints of Shanxiangyuan to detect its antioxidant active ingredients and establish a predictive model from performance indicators to its antioxidant. A data fusion method based on the Maximum Entropy model is proposed, which solves the problem that the results are not stable at multiple wavelengths. And a model averaging strategy under the two structures of nested structure and sequential structure is utilized with considering the mallows criterion, the Jackknife criterion, and the Akaike information, which not only has a good effect on the simulation data but can also reduce the prediction error of Shanxiangyuan's antioxidant value. Furthermore, this paper gives a strong correlation index of the antioxidant capacity of Shanxiangyuan according to the model averaging method.