Weighted support vector regression approach for remote healthcare monitoring

With ever increasing growth in science and technology, quality of life is improving day by day and health becomes a major concern for everyone. Since many people living in nuclear families, special measures need to be taken especially for the old age people living all alone. Wireless sensor network has become popular due to advancement in the communication capabilities of data, computational processing power and low power microelectronic devices and micro sensors. Body Area sensor network is a collection of wearable sensor nodes that implanted on human body for providing continuous monitoring of health conditions. Basically, we are focusing on how to predict the activity based on sensor reading. This research paper predicts the activity that is based on sensor reading so that we can provide continuous monitoring to patients in order to provide pervasive healthcare. It is in contrast with the traditional event driven approach where a patient visits the doctor only when he is sick. Prediction of activity is done using Support vector Regression (SVR). This paper presents a model to overcome the over-fitting which is due to noise and outliers in dataset. So, we propose an approach which uses weight factor to reduce the prediction error and result in higher accuracy than simple support vector regression.