Abstract: 【Purpose/significance】The SHAP explanation method was used to conduct interpretability analysis on the information adoption behavior prediction model of the online consultation platform to identify the important influencing factors and their modes of action that affect information adoption behavior.【Method/process】Firstly, based on the information adoption model, candidate feature variables related to the question and answer text are extracted from the dimensions of information quality and information source cred⁃ ibility, and Spearman correlation analysis is used to screen them; secondly, multiple ensemble learning classic algorithms are used to construct information adoption behavior predict the model, and use relevant performance indicators to evaluate and select it; then, use the SHAP explanation method to conduct interpretability analysis from the importance of feature variables, main effects, modes of ac⁃ tion, and interactive effects for the best overall performance model; finally, based on the analysis results, an information adoption opti⁃ mization strategy is proposed from the three dimensions of patients, doctors, and platforms.【Result/conclusion】The time difference between questions and answers and the total number of reply texts are the two most critical factors that affect patients' information adoption. In addition, factors such as follow-up questions, reward amounts, and answer text entropy also have different effects on infor⁃ mation adoption behavior in terms of degree, direction, and method.【Innovation/limitation】The algorithmic attribution idea is used to mine the influencing factors of information adoption behavior. However, the data source and form are relatively single, and the feature extraction is still limited to the question and answer record form features. In the future, the content features in the multi-source and multi-modal question and answer data will be deeply mined to improve the relevant influencing factors.