ТОЧНОСТЬ МЕТОДОВ СЛУЧАЙНЫЙ ЛЕС И МНОГОСЛОЙНЫЙ ПЕРСЕПТРОН В ЗАДАЧЕ ПРОГНОЗИРОВАНИЯ ИСХОДОВ ДЕТСКИХ ИШЕМИЧЕСКИХ ИНСУЛЬТОВ

Authors made the comparison between two methods (random forest and multilayer perceptron) to forecast the outcome of the pediatric ischemic stroke. Two options of the outcome were estimated: disability and the absence of disability. Case series included 172 patients data base, all patients had MRI confirmation of stroke and signed concern form. Eight thrombophilic genes polymorphisms: FGB:-455G>A, F2:20210G>A, F5:1691G>A, F7:10976G>A, F13:103G>T, ITGA2:807C>T, ITGB3:1565T>C, PAI-1:-675 5G>4G, and four genes polymorphisms of folic acid enzymes: MTHFR:677C>T, MTHFR:1298А>С, МТRR:66А>G, MTR:2756А>G were established as feasible predictors. Multilayer perceptron method showed higher rates of correct recognition of the outcomes, than random forest method (0,88 vs 0,67).