A novel weighed hidden markov autoregressive approach for trend prediction of electronic systems

In this paper, a novel condition trend prediction method named WHMAR for electronic systems is presented, which is based on weighed Hidden Markov model (HMM) and autoregressive model(AR). The basic idea is constructing AR prediction cells as the output of HMM, which leads to a segmentation of the time series into different AR models. The hidden state sequence of the Markov chain is chosen and predicted firstly by means of weighed method. In a second step, the output results of this model are computed by the AR model as the prediction output. The method is tested on the trend prediction of complex chaotic time series and typical electronic equipment's BIT states, and the experiment results are promising.