The invention provides a bacterial
drug resistance prediction method and
system based on a
time sequence electronic health
record, and relates to the technical field of
drug resistance prediction, and the method comprises the steps: extracting patient features from the electronic health
record; performing two-dimensional and three-dimensional
feature extraction on the
drug molecules to obtain drug features; extracting bacterial characteristics according to the bacterial
genome sequence; extracting drug sensitivity characteristics according to a drug
sensitivity test of microorganisms; respectively aligning other features with the drug sensitive features; initializing a
modal shared space, fusing the drug sensitive test
time difference into a query vector, and calculating the attention of the
shared space to each
modal feature; aggregating all the features into a
shared space according to the attention to obtain comprehensive features of the drug sensitive test; and inputting the comprehensive features into a pre-constructed prediction model to obtain a prediction result. According to the method, the problems that time features are not utilized in
drug resistance early warning, interaction between patient features is ignored, and modeling of drug molecules and
bacteria is lacked are solved, and the accuracy of
drug resistance prediction is improved.