A method for predicting the genetic risk of gestational diabetes based on artificial intelligence

Abstract

The invention discloses a method for predicting the genetic risk of gestational diabetes based on artificial intelligence, comprising the following method steps: acquiring and integrating patient physical examination data and gene detection information, establishing a medical record database of gestational diabetes; Preprocessing, including splitting the training-test set, screening medical records, and filling in vacant values; combining Information Value and Bayesian Network to extract features to construct a feature group related to the genetic risk of gestational diabetes; based on the CatBoost model, the medical record data after feature screening Modeling and diagnosis; use Grid Search to find the parameter value with the best score, and use the training set for cross-validation; the method of the invention based on artificial intelligence-assisted genetic risk prediction of gestational diabetes, combined with genetic data and physical examination data, can be very good It can be applied to the actual medical environment to find out the high-risk groups of gestational diabetes, win valuable intervention time for patients, carry out early intervention, and change the maternal and fetal outcomes.

Description

technical field [0001] The invention relates to the technical field of gestational diabetes prediction, in particular to a method for assisted prediction of genetic risk of gestational diabetes based on artificial intelligence. Background technique [0002] Gestational diabetes is one of the major diseases during pregnancy. It is a temporary form of diabetes in which the body does not produce enough insulin to regulate blood sugar during pregnancy. If gestational diabetes is not treated, it will seriously endanger the health of mothers and children, specifically manifested as increased risks of pregnancy-induced hypertension, fetal arrest, polyhydramnios, etc., and the incidence of premature birth and macrosomia will increase significantly high. [0003] At present, the traditional diagnosis of gestational diabetes is based on questionnaires of risk factors, the prediction accuracy rate is low, and the missed diagnosis rate is as high as 30-40%. Other screening methods are...

AI Technical Summary

Problems solved by technology

[0004] The existing similar patent publication number is CN109524118A, a screening method for gestational diabetes based on machine learning and physical examination data. This method proposes a screening method for gestational diabetes based on the LightGBM algorithm and physical examination data, although it also combines individual genes However, it has not processed and screened the genetic data, which has a high probability that the data will retain a lot of noise, cause the model to overfit, and cause problems such as a decline in diagnostic accuracy; Shift prediction problem, the Catboost algorithm is significantly better than LightGBM

Images