Undesirable event risk prediction method based on electronic health record of patient

The invention discloses an adverse event risk prediction method based on an electronic health record of a patient. The method comprises the following steps: S1, data preprocessing; s2, performing K-means clustering sampling processing, and dividing the data into three class clusters to obtain three clustering centers; and S3, sorting the three clustering centers from small to large according to the maximum value in P *, respectively taking the three clustering centers as a rare code subset, a relatively common code subset and a common code subset, then respectively and correspondingly inputting the three subsets into three basic classifiers of GRAM +, Dipole + and RNN + for pre-training, and then performing model fusion on the three basic classifiers. According to the method, a proper training sample is sampled for a basic learning device through a clustering algorithm, an adaptive combination strategy is designed, and integrated weights of different basic classifiers are adaptively generated according to the