Blood Pressure Prediction with Multi-cue Based RBF and LSTM Model

Blood pressure(BP) is an important indicator of one's health status. The sudden rise of BP may cause serious consequences, especially for the elderly. Early prediction of changes in BP is of great importance. However, most of the previous work did not use the user's basic information for blood pressure prediction, these works did not consider the relationship between these factors with BP. Some just added user information directly to blood pressure prediction without considering the different impact extent between these factors with BP prediction. Besides, they did not use the change characteristics of the related time series data (e.g. heart rate) for BP prediction. In this paper, we proposed a multi-factor model based on RBF and LSTM networks. The model used RBF network to process user contextual information (e.g. age, BMI, gender, temperature) which can take full advantage of all contextual information factors. Using dual-channel LSTM network structure, which can feed the prediction of the related time series data into the model together with contextual data as cues for the BP prediction. Our model made the BP prediction not only extract the characteristics of the BP data itself, but also add more relevant features. Experiments demonstrated that the proposed method achieved a better accuracy than several state-of-the-art methods.