Sample expansion-oriented LDBN heart disease risk forecast model

The risk forecast model of heart disease provides an important method for predicting the incidence of heart disease. This paper proposes a sample expansion-oriented LDBN heart disease risk forecast model, in order to solve the following problems: the difficulty of related medical data acquisition in heart disease diagnosis, less samples, the slow convergence speeds and easily caught in the local optimum of existing risk forecast models. Starting from LSTM (Long Short Term Memory) that forecast the information of the next moment with extract knowledge from existing information and combined with the current moment input information or the previous layer output information of DNN (Deep Neural Networks), the model simulates the information of patients with heart disease and adopts the corresponding strategies to expand medical data about heart disease. Then, we build a risk forecast model based on LDBN (Long Deep Belief Networks) for more accurate prediction combing with DBN (Deep Belief Network) and the improved recurrent neural Network (Long Short Term Memory) and use Heart Disease Databases of UCI to verify the accuracy of the model. Comparing with BP and DBN model, the accuracy of the proposed model is respectively increased by 2.9.