Abstract: With the development of digital transformation in healthcare, intelligently analyzing the complex and diverse diagnosis and treatment information generated during patients' diagnosis and treatment processes, identifying key features that affect diseases, and accurately predicting patients' conditions are of great significance for assisting doctors in formulating diagnosis and treatment plans and providing treatment recommendations. Based on this, this study proposes an intelligent auxiliary diagnosis method for diseases that integrates preference learning and evidential reasoning. Firstly, considering the differences in the impact of various disease features on diagnosis results, a preference learning model is established to obtain the weights of different features, thereby identifying the core feature set. Secondly, heterogeneous base classifiers are constructed based on different classification algorithms to realize the prediction of patients' conditions. Furthermore, to improve the accuracy of disease diagnosis, evidential reasoning is introduced to integrate and fuse the prediction results of the base classifiers, assisting doctors in judging patients' conditions. Finally, the feasibility and effectiveness of the method proposed in this paper are verified with actual medical datasets as examples.

Key words: feature selection, preference learning, evidential reasoning, integrated learning, intelligent diagnosis