The invention discloses a dynamic intelligent test paper composition method,
system and device based on multiple question types and multiple difficulty gradients, and relates to the technical field of online education, and the method comprises the steps: cleaning and removing duplications of questions in a question
bank, constructing question portraits containing multi-dimensional information, and calibrating difficulty; receiving test paper configuration to generate hard constraints, dividing a multi-difficulty gradient layer, and setting target difficulty and continuity constraints; the candidate set is screened according to question types and gradient
layers, and initial test papers are generated in a layered and segmented mode; the constraint conflicts are repaired through iteration replacement exchange, and
finalization is optimized under the multi-objective criterion; and finally, dynamically updating capability parameters based on answer data, adjusting subsequent target difficulty, and outputting test paper and
verification indexes. The follow-up test paper composition strategy can be adjusted in real time according to the answer condition of the examinee, the quality of the question
bank is guaranteed, it is ensured that the test paper meets hard constraints, the difficulty is smoothly increased, the test
paper quality and the evaluation effect are improved, the adaptability and accuracy of personalized testing are enhanced, and the method is suitable for various personalized testing scenes.