Research Of Uncertain Data Processing Method Based On Sparse Bayesian Learning

Recently, academia and industry are interested in research on the system of access to information from physical world, with the development of sensor technology,position technology and communication technology. For example, sensor networks and Global Position System (GPS), they gather information from physical world and some extends applications based on them. The state of temperature,humidity,pressure,dangerous gas content had been monitored by sensor nodes, which had been deployed in the natural environment. The data from sensor had been sent to the background management system, then updated the database record for user's queries. At most sensor monitoring network, data acquisition and data update with a certain period. The rate of data acquisition and data update is directly to consistency of inner world (database) and external world (physical nature).Uncertainty of the data,which in sensor network,moving objects localization,biochemistry, is the inherent properties. Uncertainty in the data due to a variety of factors, such as network transmission delay, packet loss, and the external data changes save to the database is not in real time. This is different from inaccurate data caused by measurement and calculation. Date uncertainty is the new direction in database research, meanwhile, affects all aspects production and life. So its study is of great significance.In this paper, Wuhan Iron and Steel Fiber Bragg Grating temperature monitoring sensor online network as a collection of background temperature. Temperature is affected by many factors, the collected data with uncertainty. The prediction of Relevance Vector Machine is probabilistic, which improve the defects of noise treatment. During the process, Relevance Vector Machine used fewer vectors than Support Vector Machine. Through classification and regression, the improved Sparse Bayesian learning algorithm has more excellent efficient than the original algorithm. At the same time, the collected temperature has been simulated to make a simple prediction.