Опыт разработки прогностической модели для оценки состояния репродуктивного здоровья социально-адаптированной молодежи

The aim of the study was to develop predictive model for reproductive losses risk in married couple based on medical history, clinical and laboratory data. Were surveyed 317 people from among high school youth using a specially designed questionnaire. Among high school youth hormonal research was by enzyme immunoassay with the definition of Follicle-Stimulating Hormone (FSH), luteinizing hormone (LH), estradiol (E2), testosterone (T), thyroxine (T4), thyroid stimulating hormone (TSH), thyroid peroxidase antibodies (aTPO). As a training set to build a predictive model used a sample of 48 young couples aged 18 to 35 years. 24 couples had reproductive losses in history and the 24 pairs had favorable reproductive anamnesis. Predictive classification model was made with the help of functionlibrary “randomForest” systemThen we partition into two intervals of the amounts by the k-means algorithm. The fewer predictors were produced in the model using function “importance” of the R series. The smallest number of predictors that maintains acceptable low out of bag error that are not included in the training (OOB) was 13. In this study OOB = 20%, of which 16,7% for reproductive couples, and 23,5% for nonreproductive couples. According to the results of building predictive models has been obtained the list of predictors of descending importance for the diagnosis of reproductive losses