Applicaton Of Hybrid Feature Selection Model In The Diagnosis Of Gestional Diabetes

Gestational diabetes is an abnormal condition of glucose metabolism that occurs during pregnancy,which often leads to maternal and perinatal diseases and type 2 diabetes.Early diagnosis and control of the disease can effectively reduce the risk of disease.With the development and integration of biomedical and computer science,more and more machine learning algorithms and techniques are applied to assisted medical diagnosis.At present,there are many applications of machine learning algorithms in the diagnosis of diabetes,and with the improvement of the data dimension,the feature selection model is more and more widely used in the auxiliary medical diagnosis process.This paper proposes a hierarchical feature selection model with generalization meaning,and gives a formal description,which can be used to represent the fusion of single feature selection model and multiple hybrid feature selection models.Drawing on the hierarchical structure widely used in computer science,the feature selection model is managed through a hierarchical structure to achieve model fusion in the feature selection stage.The proposed hierarchical feature selection model has theadvantages of good generalization and strong robustness,and is suitable for finding that the physical interpretation of features is not clear,and the pattern between the class tags is unknown.Secondly,this paper improves a feature selection algorithm based on association rules.According to the confidence of the rule and the frequency of the feature appearing in the rule,the weighted frequency score corresponding to the feature is calculated to represent the feature importance and realize the feature selection function.This paper uses the Tianchi diabetes dataset,which contains55-dimensional SNP features and 29-dimensional sign features.According to the characteristics of the dataset,a hybrid feature selection model is proposed,which uses the hierarchical feature selection model and the association rule-based feature selection algorithm.The feature selection of the SNP feature and the discretized feature feature is performed,and the selected features are spliced ??to finally realize the feature selection function.And through experiments,under the Tianchi diabetes dataset,the feature selection effect of the hybrid feature selection model proposed in this paper is better than other control feature selection models.