Anston attentional network for structured data based stroke risk prediction in smart aging

To reduce the pressure on public health services caused by the aging population, nursing homes need to predict disease risks for the elderly periodically. To improve the disease risks predicting ability of nursing homes, we designed Anston (An Attention Mechanism Network Model for Structured Data Classification) in the application scenario of innovative elderly care. The Anston model can use the physiological indicators and pathogenic factors easily collected by nursing homes to predict disease risks. In the study of disease risk prediction based on physiological indicators and pathogenic factors for thoughtful elderly care, we designed a data enhancement method, a feature weight automatic update method, and a multi-layer perceptron neural network to solve the problems of sample shortage, inconsistent feature weights, and sample imbalance. At the same time, we designed an attention mechanism network model for structured data classification based on the multi-layer perceptron neural network Developed in this paper. To fit the application scenario of competent elderly care, we propose a disease risk prediction model, Anston, based on the data enhancement method, feature automatic update method, and structured data classification attention mechanism network designed in this paper. We use public data sets and subject data as sample data in the experiment. The experimental results show that the Anston model has an accuracy of 95%, a precision of 92%, a recall of 91%, a specificity of 93%, an F1 score of 91%, and an AUC of 93% in predicting disease risks in the experiment, which have achieved the SOTA result.