INTRODUCTION: Chinese medicine formulas (CMF) are an important aspect of traditional Chinese medicine (TCM) and are formulated based on strict compatibility proportions guided by TCM theory. Due to the complex chemical constituents of TCM and the diversity of evaluation indicators for a certain disease, the research strategy on how to obtain the optimal combination of these crude extracts, homologous compounds or even the specific compounds mixture becomes the key step in the study of compatibility proportion research. Therefore, in this research, the "Eczema mixture" (EM) which includes six kinds of Chinese medicinal materials for the treatment of atopic dermatitis was cited as an example to illustrate the proposed compatibility optimization strategy. METHODS: Ultra-performance liquid chromatography-quadrupole/time-of-flight (UPLC-Q/TOF) technology was used to analyze the chemical components in the EM formula, and a total of 136 chemical compounds were identified. 76 formulas with different compatibility ratios were generated with the simplex centroid mixture design (SCMD). Two defined objective functions, the maximum of the anti-inflammatory and anti-allergic activity were used to evaluate the bioactivities of all the formulas. The 6-n-2 three-layers of back-propagation artificial neural network (BP-ANN) was employed to model the two defined objective functions. With the predictive models, the Pareto front was determined by a variant of non-dominated sorting genetic algorithm II(VNSGAII) to provide the optimal prescription set. RESULTS: The 6-n-2 three-layers of artificial neural networks demonstrated a satisfactory fitting effect for the nonlinear activity relationship. In the EM formula, Huangbai and Kushen were identified as the main botanical drugs with anti-inflammatory and anti-allergic roles. The results were consistent with the clinical application of the 113 prescriptions involving 230 botanical drugs for the treatment of AD from the 'Dictionary of Traditional Chinese Medicine Prescription'. CONCLUSION: The proposed SCMD-ANN-VNSGAII is a powerful approach that may facilitate future compatibility optimization of homologous compounds or specific component mixtures.