Performance Evaluation Of Crude Oil Price Volatility GARCH-class Forecasting Models Based On DEA

Crude oil that is the main chemical materials as well as the strategic national resource has great impact on the steady development of national economy. However, as a commodity trading in the global scope, it has a certain financial risk. Furthermore, influences brought by economic and other factors result in dramatic variability in the price of crude oil in the global market. Therefore, exploring the international crude oil price fluctuations plays an important role in making correct decisions of national macro policy, operations of enterprises and people's living conditions. Based on the background above, the paper tries to use the future price of WTI crude oil market as the research object and predicts its profit it could make. Because of crude oil price fluctuations will have an impact on some of the macroeconomic incidents and unknown psychological factors on investors and imperfect financial market system will make oil price fluctuate uncertainty, volatility prediction error on oil will increasing along with the major events. The community pays crucial attention to establishing a scientific mathematical models and an effective evaluation system to ccurately predict its variability. The prediction models on oil price variability are endless reseach areas and in the multiple evaluations, decision-makers are difficult to obtain a consistent model from many prediction models.In order to obtain a more consistent rank model predicted by the DEA model, this paper proposes the evaluation results of prediction models as input and output of DEA model. The details are as follows: Firstly, we establis h GARCH model and the GARCH model under long-term memory to predict price of crude oil. Studies have found that future WTI oil profit on crude oil market has long-term memory, non asymptotic and thick tail, which means crude oil price is susceptible to ext ernal factors and the other risks. Secondly, in order to get the rank of all prediction models, the paper constructs the DEA model and distincts the inputs and outputs by desirable and undesirable expectation under different decisions evaluation. In additi on, we further build the super efficiency DEA and deal with inappropriate units under these models and get a complete prediction model efficiency rank.The results are as follows: Among the GARCH models, GJR model and the GARCH model is relatively significant. GJR model is more accurate in terms of prediction errors and the direction of the oil price profit fluctuations. The EGARCH model is not as good as other models in the prediction and is better in consideration of the overall profit. In the long-term memory GARCH models, FIAPARCH model is always the best oil price prediction model This model can accurately predict the direction of oil price variability. When considering the general direction of fluctuations in profit, FIEGARCH model is always the best prediction model. Finally, the we find out that the proposed super-efficiency improved DEA model can evaluate different crude oil price prediction models, therefore it can get more consistent results ranked under different evaluations.