An improved chaotic genetic algorithm optimized LS-SVM method for economic forecasting

Accurate forecasting of some economic indicators such as GDP is very useful. Aiming at the problem of modeling and forecasting of the nonlinear and complex economic system, an improved least square support machine model is proposed in this paper. A multi-scale chaotic search algorithm combined with GA is proposed for the optimum selection of model parameters. Time series data of the indicator to be forecasted is used as the model input. Simulation results show that the prediction accuracy has been improved, the average error rate decreases from 25% by the BP neural network to less than 2% by the proposed algorithm.