FIELD: artificial intelligence.
SUBSTANCE: invention relates to a method of forming mathematical models of patients. A method comprising: obtaining a training sample containing an electronic medical history of the patients; pre-processing data contained in the patient's clinical records; converting processed data into a sequence of medical facts for each patient; performing automatic marking of the obtained sequence of medical facts on each patient; performing on server training of primary representations independently for each of modalities; training of joint representations; final model and aggregation parameters training; patient's medical history not included into the training sample is obtained; performing the patient's obtained medical history data preliminary processing; converting pre-processed data into a sequence of medical facts using medical ontologies; patient's vector presentation is formed; formed vector representation is used to diagnose, analyze and predict the development of diseases.
EFFECT: technical result consists in improvement of accuracy of diagnosing, analysis and prediction of developing diseases.
1 cl, 10 dwg
