Medium-and Long-term Forecasting Methods And Their Application

Compared with the short term forecast of things development, the middle and long term forecast of things development is usually affected by more uncertain factors. First, the article summarizes the present methods for the middle and long term forecast, and introduces the theories interrelated with those methods, briefly. Second, combining the characteristics of the middle and long term forecast, it establishes some middle and long term forecast models, excerpts the lead's yearly statistic data since the establishment of China, and makes an example application analysis to study the middle and long term forecast models. The middle and long term forecast models and methods adopted in the article are:(1) Medium and Long Term forecast based on grey-markov model. The model adequately makes use of the characteristics of grey prediction and Markov theory, forecast the fluctuation along the tendency by GM (1,1) prediction, and carries on Markov forecast based on it. The newest data is gradually added while the oldest one is removed from original data sequence.(2) Improved medium and Long Term forecast based on ANN-ARIMA model. First, this model uses the ARIMA model to forecast the long-term development trend in order to make its linear rule information contained in forecast result of the ARIMA. Second, it takes the improved BP (32) model to fit the non-linear rule based on this foundation, and obtains the forecast value of the BP(32)-ARIMA model.(3) Medium and Long Term forecast based on combination forecasting model. The paper adopts the different weight deterministic method to combine the agonic GM (1,1) model and the BP (32) nervenetwork model. The example has been discussed so as to check up the performance of different combination forecast methods in the long-term forecast performance.The article makes transverse contrast and result analysis within every model by MATLAB program software, through some examples. The example examination proves that when the relative information is inadequate, it can also satisfy the expectation and demand to the middle and long term forecast in practical work if the model was chosen properly, just making use of a certain historical period crowd and statistic method.