Multifractal-based International Oil Price Forecasting

As an important energy source,oil plays an important role in the global economy.Fluctuations in oil prices have a big impact on the macro economy.For example,the period after the early 1970s,the sharp rise in oil prices and the consequent recession have led to the view that rising oil prices are likely to lead to economic recession,excessive inflation,productivity declines,etc.Therefore,it is important to study the fluctuation of international oil prices.The oil price fluctuates drastically and the influencing factors are complex.It is not only affected by market supply and demand,but also political factors,inventory levels,psychological expectations,emergencies,strategic emergency reserves and other factors,which will make future oil price fluctuations difficult to predict.In order to improve the prediction accuracy of international oil prices,this paper attempts to construct a petroleum price forecasting system based on multi-fractal theory from the perspective of multi-fractal theory and existing oil price research methods.This paper mainly introduces three models based on the fractal characteristics of oil price:1)international oil price forecasting model based on pattern matching and fractal interpolation,2)an international oil price forecasting model based on high frequency driving from two perspectives of statistics and fractals,3)improved multi-fractal wavelet-based multi-period forecasting model for international oil prices.The specific work of the three models is as follows:(1)International oil price forecasting model based on pattern matching and fractal interpolation.The international oil price series has obvious nonlinear and long-term memory characteristics,so it can use historical pattern matching to predict the long-term memory system(i.e.oil prices),in order to overcome the shortcomings of the general short-term memory system prediction method.Based on the traditional similar pattern matching and the fractal interpolation method,this paper constructs a model framework that integrates the prediction values of multiple similar patterns to improve the prediction accuracy.Firstly,search for the matching similar pattern from the historical data;secondly,perform the same scale transformation on the subsequent sequences of the matched similar patterns to obtain the subsequent pattern;finally,consider the possibility of n(n=3 in the empirical)subsequent patterns,The extracted n subsequent patterns are integrated into a final prediction series.The empirical research shows that the proposed Hybrid ensemble prediction model is significantly better than other short memory system models.(2)International oil price forecasting model based on high frequency drive-from the perspective of statistics and fractals.The model proposes to effectively use the daily oil price data to improve the prediction accuracy of the monthly oil price series.Firstly,the multi-fractal spectrum method and the traditional statistical method are used to extract the effective fluctuation parameter information.Then,it is judged whether there is a correlation between the extracted fluctuation parameters and the monthly oil price data.Finally,adding the volatility parameter of correlation to the time series model as an exogenous variable.Studies have shown that the proposed model can improve the prediction accuracy effectively.(3)Improved-multifractal-wavelet-based multi-period forecasting model for international oil prices.The model combines the advantages of Haar wavelet and multiplicative cascade tree structure to construct an improved variant multi-fractal wavelet model(V-MWM method)to predict the international oil price in multiple periods.Firstly,the Haar wavelet three-layer decomposition is performed on the daily oil price,the coarse-grained layer(scale coefficient)data is extracted,and the scale factor is single-step predicted.Secondly,the daily oil price is decomposed into three layers with multiplicative cascade method.Extract the fine-grained(multiplier)data of each layer,and then predict the multi-period multipliers of each layer.Then,the quantitative relationship between the scale coefficients and the multipliers is constructed,and the predicted scale coefficients and predicted multipliers are used to obtain the predicted wavelet coefficients of each layer.Finally,the predicted scale coefficient and predicted wavelet coefficient are reconstructed into the original sequence granularity by Haar wavelet,and the multi-period prediction value of the daily oil price is obtained.The empirical research shows that the constructed V-MWM method reduces the computational time complexity greatly while ensuring the accuracy of prediction in the multi-period prediction of daily oil price.The three international oil price prediction models proposed in this paper analyze and multiply the multifractal characteristics of international oil price daily and monthly data and make use of them rationally.Three prediction models are constructed which are more suitable for oil price characteristics.The experimental results prove that the three prediction models proposed in this paper can improve the accuracy of oil price forecasting effectively.