Weight Of Covered Region Voting Method And Its Application In Energy Market Forecasting

In the field of forecasting,there are always more than one method to deal with a problem,and more than one institute will supply their own research on the same problem.It’s hard to say which method or information source is better and experience tells us these forecasting results are always unstable,so the research how to make full use of all the information that we have is valuable.A new data fusion method was proposed in this paper.By expanding the point estimates in the prediction results into the estimated intervals,these extended intervals were combined and the resulting union was divided into various equal-interval sub-ranges.Sub-intervals were voted and the fusion interval was selected.Through the above steps,this new fusion method introduced voting for multi-classifier information fusion in machine learning into prediction information fusion,and applied to international oil price and energy demand forecasting.This approach presented a stable and great performance.In the 7 group results of international oil price forecasts,the fusion results are always being the optimal and suboptimal.The performance of a single model is sometimes good and sometimes bad.The same stable performance of the fusion method gets in the case for the forecast of energy demand.What’s more,it doesn’t require training data,little limit of the source data,no complex computation,and it also provides a solution to combination puzzle.