A Study On Health Estimation Of Body Area Network Based On Rough Set And SVM

With the widely application of wireless sensors, wireless body area network (body area network) greatly promotes the development of intelligent monitoring system of hospital. For the reason that healthy status of the elderly living alone can be well monitored and protected, the research of healthy estimation of body region network has great practical significance.The healthy estimation of body region network is the technology designed for human physiological index.and data assessing. However, the collected physiological parameter and data of the patient are stable and unchanging in most of the period. Therefore, we only need to analyze the data and the physiological parameter in abnormal period to carry out a healthy estimation. At present, a range of traditional algorithms draws supports from anomaly detection to get the training data sets, it’s used to forecast the next moment which will be abnormal time. Then set the threshold exceptions probability by calculating the predicted value and the actual value of the residual variance, and find the abnormal successive time threshold to detect abnormalities. Although the traditional healthy estimation for BP neural network model has good results, high prediction accuracy, the parameters are often dependent on experience. Moreover, there is a large network redundancy and the number of iterations required, so it is easy to cause excessive learning and so on shortcomings. In order to solve those above problems faced by healthy Estimations, the article completes some research on improving healthy estimation algorithms, the specific contents are as follows:(1)At present, the common healthy estimation algorithms do not consider that the stability of physiological index value for each sample can be rest correspondence characterization in the same sample. We can put a plurality of normal data in abnormal period that to be characterized by a homogeneous data which can greatly reduce the data redundancy. At the same time, due to the consideration of the data in memory of the binary storage characteristics, we optimize the chain table structure of attribute reduction algorithm of rough set to achieve the reduction of information stored in memory optimization.(2) As for machine learning algorithms for healthy estimation, the number of layers and the number of BP (Back Propagation) neural network model for the hidden layer is an experience value. So it must be determined by repeated experiments, in addition that it has erratic learning and memory capacity and other shortcomings. In this paper, SVM (Support Vector Machine) learning model is introduced into the estimation algorithm. SVM healthy evaluation model can not only guarantee the accuracy of classification, but also can reduce the VC dimension of learning. It can get the global optimal solution, so the prediction generalization ability of the sample data is better than BP neural network method. At the same time, to improve the SVM kernel function via the linear combination of polynomial kernel function and the radial basis kernel function can greatly simplify the number of iterations of machine learning model, which also can promote health assessment model performance. Finally, a new healthy estimation algorithm based on SVM is proposed.