A variable weight combined forecasting method based on approximate entropy is proposed. Firstly, unlike
the traditional evaluation criterion, and considering measurement complexity of sequential sample, an optimizing model of
variable weight combined forecasting is established according to the approximate entropy of the prediction error sequential.
Then, the weight allocation problem is considered. To avoid the insufficiency of the conventional method (e.g. mean
estimation and regression analysis), online least squares support vector machine(LS-SVM) regression method is used to
achieve accurate forecasting about weight. Finally, an example shows the feasibility and effectiveness of the proposed
method.