A grey-cumulative LMS hybrid predictor with neural network based weighting for forecasting non-periodic short-term time series

This study introduced a grey-cumulative LSM predictor to compare with regression, backpropagation neural network predictor, Box-Jenkins, and Holt-Winters smoothing utilized for the applications of non-periodic short-term time series forecast. Statistics methods and neural network predictors are widely applicable on the issue of short-term forecasting. However, they encounter the crucial problem that the predicted values always cannot achieve the satisfactory results because the generalization capability in neural network predictors can perform extrapolation well. Therefore, this proposed predictor can improve generalization capability for extrapolation, and it is in fact is a hybrid model, combing a grey prediction model and a cumulative least squared linear prediction model, with the technique of backpropagation neural network based automatically adapting weights for both models. The verification of this study also experiments successfully in the stock price indexes forecast, and the results of the hybrid predictor achieved the best accuracy on the predicted stock price indexes as compared with the backpropagation neural network predictor, Box-Jenkins, and Holt-Winters smoothing.