A correction method of autoregressive model using neural network and time series prediction of traffic noise

摘要

As is well-known, the autoregressive (AR) model is very often used in the filed of time series prediction problems. The AR model provides good results only if a stochastic process is linear or nearly linear. However, for fairly and/or highly nonlinear processes, the prediction by a linear AR model may be very poor or even completely wrong. From the above practical point of view, in this report, we propose a correction method of AR model by introducing a multi-layered neural network. Finally, the effectiveness of the proposed method is experimentally confirmed by applying it to the actual road traffic noise data.