Abstract:  Stock price is affected by a large number of factors and is a typical nonstationary time series. In order to predict the stock price more accurately, a combined prediction method is proposed by combining the wavelet analysis, remanet GM (1, 1) model, and autoregressive (AR) model. By using the wavelet decomposing algorithm, the stock price is approximately decomposed into a number of signals of different scales. Then, these signals are reconstructed to form a number of low and high frequency time serials called the tendency part and random part of the stock price data. These serials are used for stock price prediction by using remanet GM (1, 1) and AR models, respectively, with respect to their different features. The predicted results of all serials are combined into the final prediction price. As shown in the experimental result obtained from an example, by the proposed method, the prediction accuracy is higher than that obtained by the traditional ones.

Key words: wavelet analysis, remanet GM (1, 1) model, autoregressive (AR) model, prediction