Varying-coefficient Model And Application Of Varying-coefficient Regression Model In Oil Price Forecasting

As a kind of basic raw materials, oil plays a very important role in the economic,politic and military, what’s more, it also plays an irreplaceable role in the field ofnational defense and national security. On account of some economic problems broughtby oil price fluctuations and influenceing people’s life, the countries of the worldbecame more and more concerned about it. So it is a current hot topic to accuratelypredict the price of oil and find a method that can avoid the risk brought by oil. It’s allknown that varying-coefficient model can get “dimension curse” around, responsespatial variation characteristics of the data. Above all, Compared with the traditionallinear model, it can response the authenticity of the data,and also has the very goodexplanatory for datas and make forecast better. All in all, it is better to choosevarying-coefficient model to predict the price of oil,which makes the oil pricefluctuations clear.First in this paper, it detailed introduces some of the basic kinds ofvarying-coefficient model,weighted least squares method and the way to deal with thedata. Then it adopts the way of utilizing logarithms for indirect factor variables ofinfluencing oil price, to build varying-coefficient regression model of no seasonal andexisting seasonal dummy variables. Meanwhile there is forecasting, comparison andanalysis for spot price sequence of West Texas intermediate (WTI) crude oil from thefirst quarter of2001to the first quarter of2012by using varying-coefficient regressionmodel from this article. The result of real evidence shows that no seasonal and dummyvariable varying-coefficient regression model built can more efficiently forecast oilprice sequence, as the same, its result is better than the result of direct factor variables.Finally, it’s reasonable to choice oil price influential factors to forecast the price bycounting the goodness-of-fit coefficient.