The human body characteristic parameters extraction and disease tendency prediction based on multi-sensing fusion algorithms

Disease prediction and the pursuit of health has always been the relish topic, especially as the aggravating population aging phenomenon. However, the existing medical knowledge used in disease prediction system is given by experts, which is vulnerable to restrictions of real-time and convenience. On the other hand, people don't want to use the high cost, complex medical monitoring system which could induce the inconvenience of everyday life. At the same time, the design and development of wearable bio-sensor systems for health monitoring has garnered lots of attention in the scientific community and the industry during the last years. In this article, we used an electronic health system based on multi-sensor fusion strategy. On the basis of not affecting the daily life, we got a variety of sensors with sensing, computational and data communication features, worn on the body and extracted the targeted health characteristic parameters of human. Therefore, we used the data mining algorithm to process the data in order to predict the disease tendency in long-term. Taking stroke prediction model as an example, after analyzing factors which are related to stroke, we used multi-sensor fusion system and collected 25 healthy attributes, coupled with a class attribute, built the training sample which contains a total of 125 cases of samples with 26 attributes. Based on the training sample, we used machine learning algorithms, where recognition rate could achieve 94.35%. On the basis of the classification weights, we got the most closely attributes associated with stroke and 9 healthy attributes combination with the best classification accuracy of 96.77%. Result shows that the system has achieved good classification accuracy and verified the feasibility of stroke prediction model proposed. Thus, electronic health system based on multi-sensor fusion has a high research value.