Prediction Model Of Serum Uric Acid Based On Multidimensional Features And Model Fusion

In recent years,with the development of artificial intelligence and big data,machine learning is more and more applied to the medical field.Our country has also put forward the direction of the development of big data.Through data mining,this paper excavates the intrinsic relationship between physiological indicators and serum uric acid from the perspective of data,which is of great significance for analyzing and studying serum uric acid,thus providing effective treatment guidance for patients and assisting doctors in diagnosis.In this paper,a prediction model of serum uric acid is proposed from two aspects: one is to extract multi-dimensional feature and fuse it,the other is to improve the traditional single model.On the one hand,the paper classifies and fuses the multi-dimensional features involved in the original data.Firstly,the original features are divided into two categories: digital features and text features.Then,we classify digital features into continuous numerical features and discrete features.For the character features,a Doc2 vec neural network model is proposed to extract the character features of the original corpus,and the character features are trained separately.Finally,the text features are combined with the processed numerical features to get a co mplete set of feature sets.On the other hand,the goal of the paper is to improve the prediction ability of the model.A single Boosting algorithm reduces the error by training a weak learner and fitting the residual continuously.This process makes the deviation decrease and the variance increase,which easily leads to over-fitting.Therefore,this paper proposes a combination of Boosting and Stacking to reduce the risk of model over-fitting by training multiple weak learners,so as to increase the robustness of the model.Finally,this paper proposes some important features for predicting serum uric acid through the model,and compares them with the existing research results in the medical field.Most of the existing achievements in medical field are based on statistics and physiological labeling,which requires follow-up examination of patients.It takes a long time,the size of the statistical population is large and the cost is high.This paper is based on a cross-sectional time,which has high efficiency,low cost and high prediction accuracy.It has important value for further study of serum uric acid.