Abstract
Semi-automatic learning of Bayesian network (BN) for inference about blood pressure (BP) was performed by using data gathered in a telemedicine experiment. In that experiment, data of medical electronic devices for home use that were placed in the participants' home were measured, collected through the Internet, and used for telemedicine. In the present study, for the aim of supporting users to judge different kinds of data like them as a whole on the client side, we tried to construct model for integrating data of 3 devices and additional data of weather statistics. 12 month data of a median per month for each subject (39 subjects) were used for graph structure learning. The resultant BN consisted of weight, number of steps, temperature, and systolic BP. Parameter estimation was performed individually for each of the 8 selected subjects. We tried to infer causally the best way of medical intervention for each subject by using those BNs, with a mind to prevention of hypertension. Different advice was obtained for different individual such as more walking and less weight.
