The study of method about diagnosis prediction based on adaptive filtering and HMM

A new method of diagnosis prediction based on adaptive filtering and HMM is proposed. Firstly, the feature extraction of time-sequence about collected diagnosis information is conducted, achieving time-sequence status information, and then on the basis of the results the future device status information vectors are obtained by means of adaptive filtering. Secondly, the HMMs for all diagnosis statuses are established and be trained according to the algorithm of Baum-Welch, meanwhile HMMs is regarded as the classifier. Lastly, the obtained time-sequence about future status are inputted into HMMs, getting the prediction results in accordance with the criterion of maximum likelihood. The experimental results prove the feasibility for this method, and its prediction accuracy is improved to some extent.