The complex chemical components of traditional Chinese medicine prescription bring great difficulty to its quality control, and the lack of standardization of quality control also restricts its international development. In this study, a new strategy combined with metabonomics based on mass spectrometry and chemometrics based on backpropagation artificial neural network (BP-ANN) was used to find quality markers in Zishen Yutai (ZSYT) preparation. The chemical analysis of ZSYT was carried out by UPLC-LTQ-Orbitrap spectrometry, and the corresponding chemical composition information was obtained by comparing with the chemical composition reported in the literature. HTR-8/SVneo cell model was used to study the pharmacological activity of ZSYT in vitro, while mouse endometrial receptivity model and premature ovarian failure model were used to investigate the in vivo activity. The BP-ANN model is then used to investigate the active substances, and then to investigate whether they can become quality markers and be quantified by standard compounds on UPLC. Finally, several active compounds are considered as potential quality markers. This proposed data-based method is suitable for discovering active compounds from herb medicine.